This page intentionally left blank

World Scientific
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Cataloging-in-Publication Data

Names: Curzon, Paul, author. | McOwan, Peter W., author.
Title: Conjuring with computation : a manual of magic and computing for beginners /
Paul Curzon, Queen Mary University of London, UK,
Peter W. McOwan, Queen Mary University of London, UK.
Description: New Jersey : World Scientific, [2023] | Includes index.
Identifiers: LCCN 2022050646 | ISBN 9789811264337 (hardcover) |
ISBN 9789811264344 (ebook for institutions) | ISBN 9789811264351 (ebook for individuals)
Subjects: LCSH: Computer science--Textbook.
Classification: LCC QA76 .C888 2023 | DDC 004--dc23/eng/20221202
LC record available at https://lccn.loc.gov/2022050646

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Copyright © 2023 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.

For any available supplementary material, please visit

https://www.worldscientific.com/worldscibooks/10.1142/13085#t=suppl

Desk Editors: Sanjay Varadharajan/Adam Binnie/Shi Ying Koe

Typeset by Stallion Press

Email: [email protected]

Printed in Singapore
This page intentionally left blank
This book is dedicated to my friend and co-author Peter W. McOwan,
who was the magical brain behind it. He originally came up with the idea
of our doing magic shows to inspire kids about computer science. He also
taught me to do magic and gave me the confidence, even as a complete
novice, to do live shows. For the book, more or less, he supplied the
magic and I fitted the computer science to it. He died before we finished
the book. It is not so much fun without him.

Paul Curzon
This page intentionally left blank
 Contents

Chapter Summaries xv

Chapter 0. Introduction 1

Part 1: Algorithmic Thinking 5

Chapter 1. Invisible Palming: What is an Algorithm? 7

Chapter 2. The Möbius Rings: Sequences 13

Chapter 3. Six Magical Objects: Selection 19

Chapter 4. Turning the Key: Repetition 25

Chapter 5. Calculator Conjuring: Programming Languages 31

Part 2: Evaluation and Logical Thinking I 39

Chapter 6. Wizardly Book Magic: Testing 41

Chapter 7. The 21-Card Trick: Logical Reasoning and Proof 53

Chapter 8. Mental Monte and the Trains of Thought:

Reasoning by Cases 63

ix
x Conjuring with Computation

Part 3: Making It Work for People 69

Chapter 9. The Cyclops’ Eye: You Cannot Trust Your Eyes 71

Chapter 10. Magically Weighted Boxes: Your Brain

Cannot be Trusted at All 75

Chapter 11. The Teleporting Robot: Keep It Simple Stupid 79

Part 4: Decomposition and Abstraction 87

Chapter 12. Pre-booked Picture Magic: Abstraction 89

Chapter 13. Trained Rice: Decomposition and Procedural

Abstraction 95

Chapter 14. The Teleporting Top Hat: Hiding Detail and

Swapping Parts 101

Part 5: Procedures and Procedural Abstraction 105

Chapter 15. Overhand Shuffle: Procedures 107

Chapter 16. Swapped Ends Overhand Shuffle: Specifying

Procedures 111

Chapter 17. Ends Overhand Shuffle: New Procedures

from Old and Layers of Abstraction 117

Chapter 18. False Top Overhand Shuffle: Parameters and

Generalisation 121

Chapter 19. Side Jogging Shuffle: Swapping One

Procedure for Another 127

Chapter 20. Cyclic False Cut: Libraries of Useful Procedures 131

 Contents xi

Part 6: Building Bigger 135

Chapter 21. The False Choice: Programming Interfaces 137
Chapter 22. The Forcing Matrix: Functions 143
Chapter 23. Forced Wizardly Book Magic: Calling
Functions — Building Bigger and Better 149
Chapter 24. Invisible Palming (Again): Sequencing
Procedures Together 153

Part 7: Abstraction and Data Representation 163

Chapter 25. Drawn to You: Codes Representing Data 165
Chapter 26. The Roman Maths Challenge: Roman Numerals 171
Chapter 27. The Lottery Trick: Place-Value Number
Representation 175
Chapter 28. Cards on Your Mind: How Computers
Represent Numbers 181
Chapter 29. On the Power of Numbers: Representing
Other Things by Numbers 187
Chapter 30. The Out-of-Body Experience:
Error-Correcting Codes 191

Part 8: Human–Computer Interaction 199

Chapter 31. The Four Aces: Usability and Attention 201
Chapter 32. Supernatural Suggestion: Visual Salience 211
Chapter 33. The Numbers Game: Visibility of System State 215
Chapter 34. The Three-Way False Cut: Conceptual
Models and Metaphors 221
xii Conjuring with Computation

Chapter 35. The Joker in the Pack: Memory 229

Chapter 36. Are You Psychic? User Experience and

Engineering Delight 237

Part 9: Evaluation and Logical Thinking II 245

Chapter 37. Are You Psychic? (Continued): Loop

Invariants and Inductive Proof 247

Chapter 38. The Red–Black Mind Meld: Abstraction,

Algebra and Proof 255

Part 10: More on Computational Thinking 265

Chapter 39. The Doomsday Clock: Generalisation 267

Chapter 40. Free My Three: Divide and Conquer 275

Chapter 41. The Double-Destination Deception:

Data Representation — Graphs and Cycles 285

Part 11: Cyber Security, Privacy and Society 295

Chapter 42. Sniff Out that Card: Steganography 297

Chapter 43. Classic Mentalism: Codes and Ciphers 303

Chapter 44. Call the Clairvoyant: Legitimate Channels 311

Chapter 45. Call the Clairvoyant a Second Time: Covert

Channels and the Surveillance Society 319

Chapter 46. Reading a Personality with Hot Chocolate:

Big Data and Privacy 325

Chapter 47. The Chevreul Pendulum: Professional Ethics 331

 Contents xiii

Part 12: Advanced Technology 337

Chapter 48. The Faerie Cage: Augmented Reality 339
Chapter 49. The Down-Under Deal and Robotic Reveal:
Robots 345
Chapter 50. The One True Sovereign: Curtain Call
and Creativity 353

Further Reading 365

Acknowledgements 369

Index of Computing Terms 371

Index of Tricks, Illusions and Conjuring Techniques 381

Index of People 385

This page intentionally left blank
 Chapter Summaries

Chapter 0. Introduction
This is a book that will teach you some magic, but also explains
the many links between computing and magic, with the result
that you will learn about both magic and computation.

Interlude: Jean-Eugène Robert-Houdin

PAUSE: Alan Turing

Part 1. Algorithmic Thinking

Algorithmic thinking is about thinking of the solutions to
problems as algorithms. We explain what an algorithm is, explain
why notation matters, and describe the three control structures
that all algorithms can be built from.

Chapter 1. Invisible Palming: What is an Algorithm?

You invisibly move a card from one pile to another.
A self-working magic trick consists of a series of steps that if
followed guarantee to achieve a magical effect. An algorithm is
also a series of steps that if followed blindly guarantee to achieve
a result.

PAUSE: Muhammad ibn Mūsā al-Khwārizmī

xv
xvi Conjuring with Computation

Chapter 2. The Möbius Rings: Sequences

You cut a ring in half but magically get two interlocked rings.
Tricks involve following a sequence of instructions. Sequences
in algorithms (and so programs) are a fundamental part of
computation. Sequencing is the first key control structure and
so a building block for describing computation.

Interlude: Felicien Trewey and Harry Blackstone Sr.

Chapter 3. Six Magical Objects: Selection

Given the choice of six magical objects, a volunteer picks the
one that you, the magician, predicted.
Branching, or selection, where something different is done
depending on the situation, is the second key control structure
that algorithms are built from.

PAUSE: George Boole

Chapter 4. Turning the Key: Repetition

A volunteer chooses a card from a pack and then places it back.
You, the magician, see neither it being chosen nor replaced, but
still can find the card.
Repetition is the final kind of control structure needed. With
sequencing, selection and repetition, you have all the building
blocks needed to describe the flow of control of any algorithm.
All computation (so tricks) can be described with just these
three control structures.

Interlude: Penn and Teller

Chapter 5. Calculator Conjuring: Programming Languages

A volunteer types a long number into a calculator based on
a shorter number they freely choose. You demonstrate your
ability to do lightning calculation by immediately predicting
 Chapter Summaries xvii

three numbers that divide exactly into it. Even more surprisingly,
when the divisions are done, the answer is the original number.
A key step in programming, and a part of computational think-
ing, is in turning a possibly imprecise version of an algorithm into
very precise steps with no room for doubt. That requires a precise
language with a very precise mathematically defined meaning.
This is a difference to magic where informal descriptions are
used.

PAUSE: Grace Hopper

Part 2. Evaluation and Logical Thinking I

We explore the need for evaluation of both programs and tricks,
and how thinking logically is at the core of computational
thinking.

Chapter 6. Wizardly Book Magic: Testing

Books about witches and wizards are suffused with magic.
A volunteer chooses a word apparently freely, and the book
then controls them, leading them to the word you predicted
in advance with the help of the book.
How can we be sure a program (or trick) always works? We need
to evaluate it for correctness. One way is exhaustive testing but
with some simple logical thinking, we can reduce the number of
tests needed.

PAUSE: Margaret Hamilton

Chapter 7. The 21-Card Trick: Logical Reasoning and Proof

A volunteer freely chooses a card from 21 laid out on the table.
No one is told which card was chosen, but you, the magician,
read the volunteer’s mind to find out which it is.
We can use logical thinking to prove a program (or trick) will
always work when the steps are followed correctly.

PAUSE: Edsger W. Dijkstra

xviii Conjuring with Computation

Chapter 8. Mental Monte and the Trains of Thought:

Reasoning by Cases
You turn your back. A volunteer picks a card from three. They
move them around. You turn back, and the three cards are mixed
some more. Despite not seeing the card being chosen nor all the
mixing, you name the card chosen.
One way mathematicians prove theorems is “by cases”. This is
really common when proving algorithms correct. Each use of
selection turns into another case to check.

Interlude: Bob Hummer

Part 3. Making It Work for People

To develop software that is easy to use, you need to understand
people, and in particular the limitations of our brains. Great
computer scientists need to understand cognitive psychology.

Chapter 9. The Cyclops’ Eye: You Cannot Trust Your Eyes

In this eye-shaped optical illusion, the centre of the eye appears
to float off the page.
Optical illusions show that you cannot believe your eyes. Just
because you see it does not mean that it is there. When designing
interfaces, you cannot assume people see what is there.

Interlude: Hajime Ōuchi and Bridget Louise Riley

Chapter 10. Magically Weighted Boxes: Your Brain Cannot

be Trusted at All
Members of the audience take it in turns to pick up a pile of
three boxes, but when, after a click of your fingers, they only
pick up two of the boxes, the pile is clearly heavier.
You cannot trust your other senses either. This illusion shows
you cannot even judge what is heavy and what is light.

Interlude: Lulu Hurst

 Chapter Summaries xix

Chapter 11. The Teleporting Robot: Keep It Simple Stupid

With this magic jigsaw, you make a whole robot appear and
disappear just by muddling the jigsaw up and putting it back
together again.
Some systems are so complicated we can’t take it all in. When
designing tricks, we aim to add complexity so the audience do not
see what is really happening. When designing usable systems, we
need to avoid overloading the senses. Keep it simple stupid!

Interlude: Sam Lloyd and William Hooper

Part 4. Decomposition and Abstraction

When inventing new tricks, we can build them out of a collection
of basic parts. Such parts can be reused across other tricks.
Likewise we can solve the problem of writing a program by
breaking it into parts to write separately, or just reusing bits
previously written.

Chapter 12. Pre-booked Picture Magic: Abstraction

An audience member randomly picks a picture from a book of
images. You are then able to correctly draw the picture chosen
despite not seeing them choose.
This trick relies on the abstraction of images. Abstraction is
about simplifying things down to their essence. Variations arise
in many disciplines. Computer scientists have many versions
including simplified images. Most versions though are both more
technical and more precise than that.

PAUSE: Susan Kare

Chapter 13. Trained Rice: Decomposition and Procedural

Abstraction
You train a bottle of rice to grip a chopstick so that the bottle
can be lifted with the chopstick, whilst the identical bottle of a
member of the audience repeatedly falls.
xx Conjuring with Computation

Even simple tricks can be described as a series of parts,

making them easier to invent and describe. This is called
decomposition. Decomposition of a problem into parts in this
way is a fundamental part of computational thinking.

PAUSE: Michael Faraday

Chapter 14. The Teleporting Top Hat: Hiding Detail and

Swapping Parts

This is a new version of Invisible Palming, but where a

teleporting top hat is used to move a card from one pile to
another.

Abstraction and decomposition allow us to think about different

parts of a program or trick at different levels of detail at different
times. This is a key use of those ideas. It also allows us to swap
in and out different parts of a trick/program to give variations,
and perhaps even better tricks/programs.

PAUSE: Jeanette Wing

Part 5. Procedures and Procedural Abstraction

There are many different kinds of false shuffles. Just by saying

the name of a shuffle, as a step in a trick, a magician would know
what was meant and be able to incorporate it into the trick.
This is procedural abstraction: a key kind of abstraction used
by computer scientists, where named, clear and self-contained
pieces of code are created for use elsewhere. Once created, they
are incorporated into the program (or “called”) by giving their
name.

Chapter 15. Overhand Shuffle: Procedures

The overhand shuffle is the standard shuffle used by card players
and magicians alike to quickly shuffle a pack of cards. Use it any
time you do really want to mix up the cards in a pack.
 Chapter Summaries xxi

The core tool for doing decomposition is the procedure. Here we

give a sub-trick and specifically a shuffle a name. That allows us
to just refer to the name in any trick we wish to use it. Likewise,
by naming program code that does a well-defined task, we can
then use the name whenever we want to refer to that code. This
is called procedure call, and it uses abstraction. We hide the
details of how the step is done, just giving the code a name. In
this and subsequent chapters, we look at shuffles as sub-tricks.

Interlude: Persi Diaconis

Chapter 16. Swapped Ends Overhand Shuffle: Specifying

Procedures
This false shuffle swaps the cards at the top and bottom of the
pack while shuffling the cards in the middle of the pack.
By giving a clear description of the effect of a trick (or program),
we do not have to worry about how it works (its implementation)
when designing other tricks (or programs) that use it.

PAUSE: Nikola Tesla

Chapter 17. Ends Overhand Shuffle: New Procedures from

Old and Layers of Abstraction
You shuffle the pack but secretly leave the top and bottom cards
of the pack alone while mixing up those in the middle of the
pack.
Once we have sub-tricks or procedures, doing well-defined things,
we can chain them together to get new ones that do something
different overall.

Interlude: Howard Thurston

Chapter 18. False Top Overhand Shuffle: Parameters and

Generalisation
This false shuffle leaves the top cards of the pack unchanged.
xxii Conjuring with Computation

Often the same sub-trick can be used in different situations but

does something slightly different. For example, it might do the
same thing but moving a different number of cards depending
on what the trick needs. This leads to the idea of parameters:
specifying the values needed to fit the procedure to the situation.

PAUSE: Ada Lovelace

Chapter 19. Side Jogging Shuffle: Swapping One Procedure

for Another
You shuffle the cards, but unknown to the audience, the top
quarter or so of the pack are unshuffled.

If we have decomposed a trick into well-defined parts, then we

can swap in equivalent parts to make variations of the same
trick, perhaps to improve the presentation of the trick. With
programs, having broken them into clear parts that do a well-
defined task, we can substitute different implementations (e.g.,
that are faster) without affecting the rest of the program. That
means we do not need to think about the rest of the program
when making such a change.

Interlude: John Nevil Maskelyne

Chapter 20. Cyclic False Cut: Libraries of Useful Procedures

You repeatedly cut the cards, but in doing so leave the cyclic
order alone.

By making collections of related techniques, like cuts and

shuffles, or in programming, procedures about a single topic, we
can organise programs so the instructions for the really useful
ones are written out once and then just called on whenever
needed.

PAUSE: Maurice Wilkes and David Wheeler

 Chapter Summaries xxiii

Part 6. Building Bigger

We have now seen most of the basic techniques that allow us
to build large programs a little at a time. In this part, we go a
step further, using results from one subroutine in the next.

Chapter 21. The False Choice: Programming Interfaces

You give a member of the audience a choice of several things,
but it is actually no choice at all.
To mix and match sub-tricks, it helps to be clear about what
they do. Procedures need to have clear interfaces, i.e., precise
descriptions of exactly what they do and when they can be
used. Interfaces help avoid mistakes being introduced when using
procedures in large programs.

PAUSE: Tommy Flowers

Chapter 22. The Forcing Matrix: Functions

Spectators choose numbers at random from a grid of numbers.
They add up to a number predicted by you, the magician.
Sometimes the aim of a particular part of a program is to return a
value — a number perhaps. That is what a Forcing Matrix does.
Its job is to come up with a number. This could be the aim in
its own right, with the magician magically having predicted the
number. It can also be used as a component of other tricks.

PAUSE: John McCarthy

Chapter 23. Forced Wizardly Book Magic: Calling

Functions — Building Bigger and Better
You first get the audience to pick a random number by choosing
numbers from a grid. Then that page in a book is taken, and
a volunteer chooses a word from it at random. The book then
leads the volunteer onto a new word in the book no one could
have known, except that you did predict it.
xxiv Conjuring with Computation

We can make more new tricks from simpler ones by chaining trick
elements together in sequences where they pass information from
one to the next. The resulting trick can be much more powerful.
The same applies to programs.

Interlude: David Devant

Chapter 24. Invisible Palming (Again): Sequencing

Procedures Together
We return to the trick where you magically move a card from
one pile to another.

Procedures with parameters allow us to create generalised

programs. We can decompose them into parameterised parts and
sequence them one after the other, using the same parameters.
By matching preconditions and postconditions, we can give an
outline argument as to why the whole algorithm works.

PAUSE: Robert W. Floyd and Tony Hoare

Part 7. Abstraction and Data Representation

We’ve seen abstraction applied to control structures in the form
of procedures. Now we explore similar ideas with data: data can
have different representations with that representation hidden.

Chapter 25. Drawn to You: Codes Representing Data

Several volunteers each draw a picture of their choice. Just by
looking at the drawings, you are able to work out who drew
which.

Tricks often rely on representing information in some particular

way. Choosing an appropriate data representation is core to
computational thinking and programming.

Interlude: Joseph Jastrow

 Chapter Summaries xxv

Chapter 26. The Roman Maths Challenge: Roman

Numerals
This is a challenge rather than a magic trick: can anyone make
a Roman sum add up correctly?
Roman numerals, as used here, are an early example of a number
representation different to the one we use. Computers use a
different one again. There are many ways to represent the same
data.

Interlude: Nevil Maskelyne

Chapter 27. The Lottery Trick: Place–Value Number

Representation
You turn the room into a gigantic lottery machine, bouncing
lottery balls made of screwed up paper around the room. When
a series of balls are chosen, they give a number that does not
match anyone’s lottery number . . . except yours.
We use the Hindu–Arabic numeral system. The really amazing
idea is to give digits a position in a number and have their
value change depending on the position, so 2 can stand for
2, 20 or 200 depending on its position. This makes large
numbers more concise but also makes operations like addition
and multiplication easy: they make for simple algorithms.

PAUSE: Charles Babbage

Chapter 28. Cards on Your Mind: How Computers

Represent Numbers
Four members of the audience jointly pick a number and
concentrate on it helped by a magic set of cards. You can
immediately read their collective mind and tell them the number
instantly.
This trick uses the same representation of numbers as the one
computers actually use: binary.

PAUSE: Gottfried Leibniz

xxvi Conjuring with Computation

Chapter 29. On the Power of Numbers: Representing Other

Things by Numbers
A volunteer picks a postcard from a selection. Your partner
magician, who is out of the room at the time, can immediately
say which card was touched.
All sorts of things can be represented by numbers, and once they
are, we can bring them into the digital world, as those numbers
in turn can be represented by binary.

Interlude: Luca Pacioli

Chapter 30. The Out-of-Body Experience: Error-Correcting

Codes
A grid of face-up and face-down cards is set up at the front
of the room by a volunteer. Blindfolded at the back the whole
time, you can still later say which card was turned over.
A more advanced data representation is that of the error-
correcting code. This kind of data representation allows mistakes
to be detected and even corrected automatically.

Interlude: Tim Bell

Part 8. Human–Computer Interaction

We turn to the presentation side of tricks and the importance of
understanding people (i.e., cognitive psychology) in more depth.
In computing terms, this is about usability and user experience.

Chapter 31. The Four Aces: Usability and Attention

You deal out hands of cards containing Aces. From the way they
have landed, one in each pile, everyone can see that on the next
deal they will all end up in the volunteer’s hand. Somehow you
manage to steal them without anyone seeing to get the perfect
hand yourself.
 Chapter Summaries xxvii

Magicians control your attention, making you miss things that

would give the trick away. When designing the interface of an
interactive computer system, you need to control their attention
too, but making sure the user looks in the right place and so
sees everything that is important.

Interlude: Gustav Kuhn

Chapter 32. Supernatural Suggestion: Visual Salience

You show the audience five cards on the screen, asking everyone
to focus their attention on just one of their choice. You, however,
have controlled them all, and when you remove a card, it turns
out everyone was focussing on that same card.

What you see depends on where you focus your attention, and
where you focus your attention depends partly on how salient
the things in the scene are. If you want important things to be
seen on a screen, then you have to make them visually salient.

PAUSE: Peter W. McOwan

Chapter 33. The Numbers Game: Visibility of System State

A member of the audience calls out 10 random numbers that
you write on different pieces of paper. A volunteer picks one at
random. Amazingly, you tell them the number on the piece of
paper they picked.

A key to a good trick is that the audience lose track of actually

what is happening. They think the state of the magical system
is one thing when actually it is something else. Visibility of
the system state is an important interaction design principle
in software. Users need sufficient feedback to track the internal
state of the system. Planes have crashed when this has been
done badly and software has behaved like a trick!

PAUSE: Don Norman and Jakob Nielsen

xxviii Conjuring with Computation

Chapter 34. The Three-Way False Cut: Conceptual Models

and Metaphors
You cut the cards twice and then reform the pack. Unknown to
the audience, the pack is completely unchanged from before the
cuts.
A good design needs a clear conceptual model that leaves the
user in no doubt about what is happening. A good magic
trick does the opposite obfuscating what is happening, so the
audience think something else is happening.

Interlude: Jay Ose

Chapter 35. The Joker in the Pack: Memory

A volunteer thinks of a number and deals out that number of
cards. In doing so, they have brought the only joker in the pack
to the top of the deck.
Our working memory is very limited. Give it too much to
remember and we start to forget things. Magicians rely on this.
Interaction Designers must avoid it.

PAUSE: The Xerox PARC Team

Chapter 36. Are You Psychic? User Experience and

Engineering Delight
You test a member of the audience for superpowers. Can they,
without seeing any of the values on the cards, end up with a
matching pair?
Usability is about whether a system is easy to use. Just as
a magician wants more than a trick just working, appearing
magical, interaction designers also go further than just ease of
use. Magicians show how User Experience can be designed into
a system.

Interlude: Fay Presto

 Chapter Summaries xxix

Part 9. Evaluation and Logical Thinking II

In this part, we look at more advanced logical thinking including
real-world (and magical-world) uses of algebra.

Chapter 37. Are You Psychic? (Continued): Loop Invariants

and Inductive Proof
How can we be sure that our volunteer will have the apparent
superpowers so that the previous trick definitely will work?
The idea of a loop invariant provides the basis of logical
reasoning about repetition.

Interlude: Adelaide Herrmann

Chapter 38. The Red–Black Mind Meld: Abstraction,

Algebra and Proof
You are able to meld your mind with that of a volunteer,
controlling their choice of cards to make sure a property you
predict holds once all the cards are dealt.
Rigorous argument is one way to convince ourselves that an
algorithm (or trick) works, but we can go a step further and use
algebra. We describe the algorithm with maths, do some algebra
on the result, and so prove desirable properties hold. This is a
vital way to check safety-critical software is correct.

PAUSE: Shafi Goldwasser

Part 10. More on Computational Thinking

We now look at some more advanced forms and uses of
computational thinking: advanced versions of generalisation,
decomposition and data representation.

Chapter 39. The Doomsday Clock: Generalisation

The doomsday clock predicts the most important hour in a
person’s life. A volunteer has a free choice of a starting number,
but still the clock stops at the hour predicted for them.
xxx Conjuring with Computation

Having worked out a principle (the mathematics underlying a

trick), we can use it to make a more general principle that
can then be used to design apparently different tricks that
underneath rely on the same maths. We can generalise code
in a similar way, allowing it to be used in wider situations.

PAUSE: John von Neumann

Chapter 40. Free My Three: Divide and Conquer

Three cards of a kind, chosen by a volunteer, are lost in the
pack. You discard cards repeatedly with no sign of them until
eventually they turn up . . . the very last cards to be revealed.
Decomposition involves breaking a problem into one or more
smaller problems that are easier to solve. Divide and conquer
is a powerful version of this where these smaller problems are
the same problem as the original (just smaller). Those smaller
problems are solved in the same way. This gives a way to create
very fast algorithms.

Interlude: Alex Elmsley

Chapter 41. The Double-Destination Deception: Data

Representation — Graphs and Cycles
A volunteer takes a series of tickets to plan a trip of a lifetime
crisscrossing the world. Despite their free choice of route, once
done you reveal that you bought in advance precisely the ticket
they need to get back from their final destination to their start
point.
An important data structure is the graph, and one of the most
important problems in computer science involves creating a cycle
(i.e., a round trip) through a graph. A famous puzzle about
the bridges of Königsberg kick-started the whole idea of graph
representations.

Interlude: Dynamo
 Chapter Summaries xxxi

Part 11. Cyber Security, Privacy and Society

There are overlaps between magic tricks and concepts from
cybersecurity. Many magic tricks are about the magician having
information the audience think they cannot possibly have
obtained. Cybersecurity and privacy are concerned with how to
prevent such information leakage.

Chapter 42. Sniff Out that Card: Steganography

A volunteer chooses a card and holds it to their chest. They hide
it back in the pack, but you can sniff it out from the faint trace
of their scent left on it because they held it close.
Steganography is a way of hiding secret messages in apparently
innocuous things. What looks like a normal message or image
disguises the real thing being communicated. The same idea can
be used as the basis of tricks to allow a magician to know the
information he or she could apparently not possibly know.

PAUSE: Sir Francis Bacon

Chapter 43. Classic Mentalism: Codes and Ciphers

Your partner magician leaves the room while a card is chosen.
Despite not being there at the time, when they return, they can
say exactly which card was chosen.
Classic mentalism tricks involve secret codes between partici-
pants. Similar codes based on code books have been the basis
of algorithms for sharing secret messages for as long as people
have wanted to share secrets.

Interlude: Reginald Scot

Chapter 44. Call the Clairvoyant: Legitimate Channels

A card is chosen from a pack. You phone your friend, the
mysterious “Clairvoyant”, and when you ask them to do so,
they tell the volunteer which card was chosen.
xxxii Conjuring with Computation

Computer systems are kept secure by having separate areas with

different security levels. Nothing from a high-security area should
be able to leak to a lower-security area. Hiding information in a
legitimate channel (an allowed way to communicate) is one-
way information might be leaked. Hackers exploit legitimate
channels in computer systems. Magicians build tricks around
leaking information via legitimate channels too.

Interlude: Sydney and Lesley Piddington

Chapter 45: Call the Clairvoyant a Second Time: Covert

Channels and the Surveillance Society
An object from a table of objects is chosen, and the mysterious
“Clairvoyant” is called. They are put straight onto speaker
phone, so everything can be heard by all. They can name the
object chosen.
Sometimes information is leaked via a legitimate channel. Another
way is to set up a covert channel: a way of communicating that is
not supposed to be there. Hackers exploit covert channels in
computer systems. Magicians also build tricks around them.

PAUSE: Hedy Lamarr

Chapter 46. Reading a Personality with Hot Chocolate: Big

Data and Privacy
After stirring it three times, a volunteer drinks a small cup of
hot chocolate you make them. You read their personality in the
sludge left in the bottom of the cup.
Horoscopes and some personality tests rely on a kind of fakery. Big
Tech companies rely on a different kind of fake personality test,
gathering as much data as they can about those who use their
services. They then make judgements about what kind of person
they are and sell the information to advertisers or political parties
who exploit this intimate “knowledge” of you to sway opinions.

Interlude: Phineas T. Barnum

 Chapter Summaries xxxiii

Chapter 47. The Chevreul Pendulum: Professional Ethics

A volunteer holds a pendulum made from an old magical key.
You ask them questions, and the pendulum, controlled by the
spirit world, swings to give the answers.
People mainly use magic for entertainment, but some use it
unethically to con people. Likewise, computing skills can be
used ethically or unethically. Many magicians debunk frauds by
exposing the tricks they are using. Some hackers, called white
hat hackers, do a similar thing. They are employed by companies
to attack the company’s own systems, to protect them against
black hat hackers, who can, intentionally or otherwise, cause
lots of damage. All professional computer scientists are required
to study ethics.

Interlude: James Randi

Part 12. Advanced Technology

Many magic tricks rely on technology to provide the effects.
Here are just a few ways such computing-linked technology has
formed the basis of tricks.

Chapter 48. The Faerie Cage: Augmented Reality

You reveal to the audience that faeries are not only real but you
have trapped one in your special rowan-lined box.
A Victorian invention for theatre, Pepper’s Ghost, forms the
basis of both illusions and early versions of augmented reality
where the real and virtual worlds mix.

Interlude: John Henry Pepper and Henry Dircks

Chapter 49. The Down-Under Deal and Robotic Reveal:

Robots
You do the special down-under deal, ending with a card on the
table that no one could have predicted. Despite that, as the
xxxiv Conjuring with Computation

card is shown to the audience, your helper, a robot, appears

from down under the table to reveal that it knew what the card
would be.
The final reveal can make a trick extra special. Automata, the
precursors of robots, have long been used as props for tricks.
Here a robot is revealed to have magically forced the card that
was finally selected. Robots are now being used in workplaces
and homes, in dangerous environments to make them safe, and
as killers.

PAUSE: Ismail al-Jazari

Chapter 50. The One True Sovereign: Curtain Call and

Creativity
Who is the one true King or Queen of Britain? You have found
a long-lost magical artefact that can reveal the answer.
We finish with a summary of the main links between conjuring
and computing and reveal one last trick (about authentication).

Interlude: Richard Garriott

 Chapter 0

Introduction

This book is for people who want to learn to do magic tricks for others
but who are also interested in learning more about computation and
how ideas from computer science overlap with those of conjuring.
Computer science and magic have a lot in common. Famously,
Arthur C. Clarke said that sufficiently advanced technology was like
magic. In fact, magicians have always known this and often work at
the leading edge of science and engineering to invent new tricks. As
it is at the forefront of technology, understanding computer science is
important for magicians interested in technology-led magic. The link to
computer science is much deeper than just using advanced technology,
though.
A popular saying is that Computer Science is no more about
computers than astronomy is about telescopes. It is at core about com-
putation, which is what computers do. What is computation? A simple
version is that it is just doing calculation by following instructions.
A more sophisticated version is that it is about the manipulation of
symbols. Conjuring is also closely linked to computation.
Computer scientists also learn to think in a particular way, called
computational thinking. It is the general skill that underpins program-
ming but is also important in more general situations. Magicians, it turns
out, have to think in a very similar way to computer scientists.
Computational thinking is in part about creating precise instructions
that always achieve a desired effect. When applied to magic, it is not
about the performance of tricks. That is playing the part of the computer
following the instructions in a program to get something done. It is more

1
2 Conjuring with Computation

about inventing new tricks, modifying old ones to improve them, and
finding ways to be sure they always work. Many more general ideas from
computer science have close links to magic too. We can learn a lot about
computer science from magic, and vice versa, and that is what this book
is about.
All the tricks in this book are simple ones that are fairly easy to
perform, and variations of which are widely known and easily found on
the Internet or in introductory magic books. Just as computer science
is about creating new computer systems, we hope that they will inspire
you to start to invent your own tricks. Perhaps at first, you might just
make tweaks to these tricks, changing the presentation, or combining
effects, but you will ultimately move on to invent your own completely
new tricks.
When you do the tricks for others as part of a magic show though,
keep the secrets. Never use them for anything other than entertainment
(or if you are a teacher, for education).
The best magic, by professional magicians, of course still leaves us
as baffled as everyone else.
 Introduction 3

Interlude: Jean-Eugène Robert-Houdin

Jean-Eugène Robert-Houdin (1805–1871) became a clockmaker
after failing in his studies to be a lawyer because he was constantly
distracted by mechanical gadgets. Magic started off as a hobby, but
he went on to become the greatest magician of his time, pioneering
the idea that magic shows should be the height of elegance. He
even took over a theatre to perform in. However, in his first show,
there was a disaster as he suffered from stage fright and came close
to having a breakdown. He was about to abandon the whole idea
but then was goaded by a friend telling him it had been a silly idea
anyway. That made him determined to prove the friend wrong. His
confidence returned, and as a result of carrying on, he completely
changed the way magic was both performed and perceived. It
went from being seen as a seedy fairground act to high-prestige
entertainment. He combined science, engineering and creativity,
inventing many gadget-based tricks, and making much use of new
scientific discoveries such as electricity and electromagnetism in
doing so. He built and used automata in his tricks, some that were
magical in their own right such as a mechanical figure that could
write. Others were combined with illusions, such as his Marvellous
Orange Tree, a machine that apparently grew real oranges, the last
one containing an item from the audience that had previously been
made to disappear.
4 Conjuring with Computation

PAUSE: Alan Turing

Arguably the most influential computer scientist ever, Alan Turing
(1912–1954), set out what it meant to do computation and proved
that not everything could be computed, all before any working
computer actually existed. His idea of a Turing Machine was a
version of a computer simplified to its essence before any computer
had been made. He proved that it could do any computation that
any computer could possibly do. In the Second World War, he
was a key person in the team that cracked the German ciphers,
allowing the Allies to read their secret communications. He worked
on machines that were the precursors to the first computers and
after the war designed a computer himself. He was one of the
earliest people to work on the idea of Artificial Intelligence, devising
a test, now called the Turing Test, that suggests we should consider
a machine intelligent if we cannot tell it apart from a human by
questioning both. He sadly died young, committing suicide by eating
a poisoned apple because of his persecution for being gay by the
British authorities. He had been prosecuted and then “chemically
castrated”. He was officially pardoned in 2013. Imagine what more
he would have achieved, and we lost, had the world been a more
tolerant place.
 Part 1

Algorithmic Thinking

Computers cannot think for themselves. They just blindly follow instruc-
tions. Those instructions must always work when followed. Algorithmic
Thinking is the skill of writing such instructions.
This page intentionally left blank
 Chapter 1

Invisible Palming
What is an Algorithm?

Conjuring
You invisibly move a card from one pile to another. You can do it
even as a novice magician who has no idea how it is done.

Computation
An algorithm is a set of instructions that if followed precisely, and
in the right order, without thought, leads to a guaranteed result.

The Trick
You need 15 identical cards. One way to get them is to buy several
identical packs of playing cards, then take out and use the Jokers so all
15 cards are the same. Have a volunteer put out both hands as though
they are playing the piano with fingers and thumbs touching the table
(see Figure 1.1). Tell the audience they must chant the magic words:
“Two cards make a pair” each time you take a pair of cards.
Take two cards and, as everyone says “Two cards make a pair,”
place them together between a pair of fingers or finger and thumb of

7
8 Conjuring with Computation

Figure 1.1. In Invisible Palming, the magician places pairs of cards between the
fingers/thumb of each hand of the volunteer, with the last place taken by a single
card.

the volunteer. Keep doing this until you have one card left. Place it
between the final fingers saying there is “one leftover”.
Now take the first pair back, again all chanting “Two cards make a
pair”. Separate them and place them face down on the table to start two
piles, one card from the pair in each pile. Take each pair back in turn
adding them to the piles: one card from each pair being neatly placed on
each pile. As you do, everyone should continue to say the magic words.
Eventually, only the final, single card is left. Take this card saying: “We
have one extra card.” Let the volunteer place it on top of one of the
piles of their choice. Square up the piles pointing out: “That pile now
has the extra card.”
Explain that now the room is full of magic from all the chanting,
you are going to do “Invisible Palming”. The extra card is on one
pile. You are going to invisibly move it to the other. Place your hand
over the pile with the extra card. Rub the back of your hand to
“make the card go invisible”. Lift your palm showing that the card
is there but is invisible. Move your hand to the other pile. Tap it,
“to make the card drop”. Announce that the card has now moved
piles.
 Invisible Palming: What is an Algorithm? 9

To show the magic worked, take the pile where the extra card was
placed and count off a pair at a time into a new single face-down pile.
Everyone still should be saying the magic words as you take each pair:
“Two cards make a pair. Two cards make a pair. . .”. Make sure the pile
is neat. You find there are only pairs: the extra card has gone! So where
is it? Take the other pile and do the same, peeling off pairs and putting
them back into a pile. Amazingly the extra card is there. It has moved
from one pile to the other!
Before thinking about how the trick works, do it, just following the
steps with a friend as the volunteer (you may find the crib sheet of steps
below helpful: see The Magical Algorithm below). If you follow the steps
precisely, it should just work . . . even if you have no idea how you did it.
You could even swap with your friend and let them be the magician. It
should work for them too.
How can you do a magic trick, magically moving a card from one
pile to another without knowing what you did or how it works?

The Magical Algorithm

Here is the trick written out as a series of steps in a structured way
similar to the way that computer scientists write out instructions.
TO DO Invisible Palming:
1. Take 15 identical cards.
2. Ask a volunteer to place their hands on the table with fingers
arched as though playing the piano.
3. DO THE FOLLOWING 7 TIMES:
a. Say “Two cards make a pair”.
b. Place 2 cards between two of the volunteer’s fingers or finger
and thumb.
4. Place the last single card between the remaining fingers or finger
and thumb.
5. DO THE FOLLOWING 7 TIMES:
a. Say “Two cards make a pair”.
b. Take a pair of cards from the fingers and split them placing
them one in each of two piles built up on the table.
6. Have the volunteer place the last card on a pile of their choice.
10 Conjuring with Computation

7. Place your hand over that pile, rubbing the back of your hand.
8. Lift your hand, show the palm pointing out that the card is there
but invisible, and then place it over the other pile.
9. Tap your hand and remove it, saying the card has moved.
10. Pick up the first pile.
11. DO THE FOLLOWING 4 TIMES:
a. Add 2 cards from the pile you are holding into a single, neat
pile on the table.
b. Say “Two cards make a pair”.
12. Note that the extra card is no longer there.
13. Pick up the second pile.
14. DO THE FOLLOWING 3 TIMES:
a. Add 2 cards from the pile you are holding to a pile on the
table.
b. Say “Two cards make a pair”.
15. Point out that you are holding the extra card: it had moved to
the second pile.

How It Works
Magicians call this kind of trick a self-working trick. It always works if
you follow the steps precisely. It works whatever the volunteer does given
the choices they have (so here wherever the extra card is placed).
There are 15 cards. After dealing out the pairs into the two piles,
those two piles are identical: there are 7 cards in each. Of course, you
don’t give anyone a chance to count the cards. First, think about the
pile where you put the last, supposedly “extra”, card. That card makes
the pile up to 8 cards: 4 pairs. When you count out in pairs, there will
only be pairs there, so apparently no “extra”card anymore. It appears to
have vanished, but it just finished a pair where it was.
The other pile will be left with 7 cards: 3 pairs with one left over.
When you count in pairs, that one left over seems to be the original
extra card. It seems to have appeared there, but it was there all the time.
You pretend it has magically moved without doing anything. Nothing
has to move!
It appears magical only because you have confused everyone. They
believe when they add the last card to a pile that they are adding an
 Invisible Palming: What is an Algorithm? 11

extra odd card. You are actually completing the last pair, though. It
doesn’t matter which pile they add the last card to: that card will turn
an odd pile into an even pile. You count out the cards in pairs to hide
the exact number.

Thinking Computationally
Algorithms, magic and programs
Computer scientists have a similar idea to self-working tricks called
algorithms. An algorithm is just a set of precise instructions that if
followed exactly leads to a guaranteed effect. The instructions have to
cover all eventualities, just as a trick does.
Once you have created an algorithm that works, you don’t have
to think about solving the original problem anymore or what to do to
achieve that effect. You just blindly follow the algorithm, and the right
thing happens. It is just like following the Invisible Palming instructions:
the magical effect of apparently invisibly moving a card just happens.
If you followed the steps exactly, it will have worked for you even if
you had no idea how it worked. In doing the trick by following the
instructions, you were acting like a computer following an algorithm.
More specifically you were acting as a computational agent. It is
something or someone following algorithmic instructions precisely and
blindly.
A computer cannot think for itself. It can only blindly follow
instructions (the instructions are its program). They were written
for it by a programmer. It does computation as a result. Computers
are therefore computational agents. Computer programs are essentially
algorithms written in a programming language (like Python, Java or
Scratch). They are languages of instructions in a form that a computer
can follow.
The detail matters in algorithms. Take our trick. Just saying we
should pretend to move a card from one pile to the other isn’t good
enough. We need to say exactly how to do it in a way that cannot be
confused. Similarly, a program needs to tell the computer exactly how
to do every little step.
12 Conjuring with Computation

Wizards
A magician who invents new self-working tricks is creating algorithms.
A programmer writing a program is doing the same. They are both using
algorithmic thinking skills: working out a series of steps that will always
have the effect they are after. Algorithmic thinking also involves writing
instructions really clearly and precisely so that there is no confusion
about what to do as the instructions have to work when being followed
by a Computational Agent.
Expert programmers, male or female, are often called wizards. They
really are!

PAUSE: Muhammad ibn Mūsā

al-Khwārizmī
The word algorithm derives from the name of Muḥammad ibn Mūsā
al-Khwārizmī (c.780–850). This was because of his book, On the
Calculation with Hindu Numerals which introduced the place-based
number system (of 1s, 10s, 100s, 1000s, etc.) to the west that we
use today. He described the basic algorithms needed to do arithmetic
using decimal numbers on a sand board: a tray of sand that made it
easy to write and rub out numbers. Eventually the algorithms from
his book removed the need for the abacus to do arithmetic.
 Chapter 2

The Möbius Rings

Sequences

Conjuring
You cut a ring into two pieces down the middle, only to find you
still only have one ring. Try again, and at last, you have two rings,
but bizarrely they are linked together.

Computation
There are three fundamental ways to organise instructions. The
simplest way is sequencing: following instructions one after the
other in the given order.

The Trick
Start with a strip of paper about 5 cm wide and at least 30 cm long
(ideally longer) and glue it in a loop, but with a twist so the top of one
end of the strip is glued to the top of the other end (see Figure 2.1). This
is called a Möbius strip. Show it to the audience without mentioning the
twist. Explain that you will cut the loop into two thinner rings. Gradually
cut it down its centre. When you get back to the start, confidently say
you now have two rings. Gently untangle them.

13
14 Conjuring with Computation

Figure 2.1. A Möbius strip. Cut along the dotted line all the way around the centre
to make two interlocked rings.

Look shocked when you see it is still a single ring and wonder out
loud how that could possibly happen, saying that you will try again.
Take this new longer, thinner ring and carefully cut it down the middle.
Say you must have two rings now, and gently untangle them. Look even
more shocked as you show that you do have two rings, but somehow
they became interlinked as you cut them.
Tell the audience that there is clearly far too much magic in the
room if paper rings are going to do that to you.

The Magical Algorithm

TO DO the Möbius Rings:
1. Create a Möbius strip by glueing a strip of paper into a ring with
a twist in it so the top of one end is glued to the top of the other
end.
2. Say you are going to cut it into two thinner rings.
3. Cut it down the middle.
4. Shocked, show the audience that somehow it is still just a single
(longer) ring.
5. Try again, cutting it down the centre.
6. Even more shocked, show the audience that it is now two rings,
but somehow they are interlocked.
 The Möbius Rings: Sequences 15

How It Works
This is another simple self-working trick that relies on an area of maths
called topology. Cut a Möbius strip down the middle, and it makes a
longer strip. Do it again, and you get two interconnected rings. It is all
due to the twist which meant you had created a ring with only one side:
draw a pencil line down the middle and it will eventually join back to
the start, having covered both sides of the original strip!

Thinking Computationally
The algorithm behind this trick is about as simple as can be as it is
only a few steps. There are two aspects of an algorithm that matter: (1)
the actual instructions and (2) the order you do them in. The latter is
called the flow of control. It is the plumbing that connects instructions
together.
This algorithm is just a sequence of instructions that must be
followed one after the other, in the order given. Follow each step
correctly, exactly once and in the right order . . . and the magic will
happen. This way of linking instructions is called sequencing. It is the
simplest kind of control structure: constructs that determine the flow
of control, the order things are done.
Sequencing of individual commands is the core to other kinds of
instructions that people and computers follow including recipes, the way
you learnt to add numbers, and programs.
Simple programs are very similar to our magic trick’s description.
They are a sequence of instructions to be followed one after the other in
the given order (see Figure 2.2). Here, we have labelled the steps (1) (2)
(3) . . . to show the order to do them. Most programming languages just
use the order that the instructions are written on the page to indicate
the correct sequence, so in most languages the steps tend not to be
numbered.
Any computer or computerised gadget that is doing one thing after
another is following a sequence of instructions some programmer previ-
ously wrote. Take a chip-and-pin card machine. It shows you the amount
you owe, asks you to touch your payment device (card, phone, etc.) and
16 Conjuring with Computation

Figure 2.2. The sequence of instructions for the Möbius rings trick as a flowchart.
The instructions are followed one after another.

then tells you it’s paid: it is following a sequence of instructions in a

program to do that.
There are other important forms of flow of control that we use
when writing programs (and planning magic tricks). Perhaps, surprisingly
though, with only two more kinds of flow of control (described in the
next two chapters), it is possible to specify the flow of control of any
computation.
 The Möbius Rings: Sequences 17

Interlude: Felicien Trewey and

Harry Blackstone Sr.
The Möbius rings trick was first performed as a magic trick, rather
than just a science demonstration, by Felicien Trewey (1848–1920)
at the end of the 19th century. Variations of the trick, which are often
called the Afghan Bands, have been performed by many magicians
since, including Harry Blackstone Sr. (1885–1965). He was one of
the great magicians of the 20th century, not only inventing tricks
but devising clever variations of existing tricks. He often introduced
modern technology twists, such as sawing a woman in half, but in his
case with a circular saw. One of his most famous illusions involved
a lit electric bulb that, in a darkened theatre, floated away from its
lamp, through a hoop, and out over the audience, all while remaining
lit. It was designed and built for him by the famous inventor of
electronic gadgets, Thomas Edison. It was the first magic trick ever
to be considered worth accepting by the Smithsonian museums.
This page intentionally left blank
 Chapter 3

Six Magical Objects

Selection

Conjuring
A volunteer chooses one from a series of magical objects, only to
find that you predicted their choice.

Computation
The second way of determining which instruction to follow next is
called selection: choosing which of several sets of instructions to
follow next based on the truth or otherwise of some test.

The Trick
Before the trick, slide a note into a pack of cards with the message “You
WILL choose ME” written on it. Place six different objects, including
the pack of cards, in a line on the table. All should have some magical
connection. You might, for example, use a magic wand, a cup with a ball
underneath, a pack of cards in their box, a silk handkerchief, a coin and a
watch. Ideally, all should be solid objects that could not possibly contain
a note, except for the cards. The cards should be in the third position

19
20 Conjuring with Computation

from the left. You stand on one side of the objects with a spectator on
the other.
Explain that magical objects get infused with magic if they are used
in magic shows and all these objects are especially magical because they
have all been used in tricks many times.
Ask the volunteer to choose a number from 1 to 6 telling everyone
what it is.
The number they choose, when you count to that object, leads to
the pack of cards. Explain that as you do lots of card tricks the cards
are full of magic and like to be chosen. Ask them to open the pack of
cards and see what is inside, then get them to read the note. Note again
how magical your cards are.

How It Works
You do something different depending on their choice of number (see
Figure 3.1). Whichever number they said gives a different way of
counting that forces the choice of the third object: the pack of cards.
If they choose the 1, then you start at your left and count out letters
to the third object: O . . . N . . . E pointing to an object on each step.
If they choose 2, then similarly you count to the third object:
T . . . W . . . O.
If they choose 3, then you just count to the third object: 1 . . . 2 . . . 3.
If they choose 4, then ask them to count out 1 . . . 2 . . . 3 . . . 4 from
their left (your right) to end on the 3rd object from your left (the 4th
from theirs).
If they choose 5, then ask them to say the letters of the word FIVE
from their left (your right) F . . . I . . . V . . . E to end on the 3rd object
from your left (4th object from theirs).
If they choose 6, then you count out the letters to the third object:
S . . . I . . . X.
They have been given a false choice: you did not tell them in advance
what you would do with the number they picked. As you only do the
trick once, you only do one of the alternatives, so they do not get any
inkling that it would have happened differently another time.
 Six Magical Objects: Selection 21

Figure 3.1. Whichever number is chosen, you choose a counting method that counts
to the third object.

The Magical Algorithm

TO DO Six Magical Objects:
1. Place a note inside a pack of cards saying “You WILL choose ME”.
2. Lay out six objects in a line with the pack of cards at position 3
from your left.
3. Ask a volunteer to tell you a number from 1 to 6.
4. IF they say 1 THEN
a. You count to the third object: O . . . N . . . E.
OTHERWISE IF they say 2 THEN
b. You count to the third object: T . . . W . . . O.
OTHERWISE IF they say 3 THEN
c. You count to the third object: 1 . . . 2 . . . 3.
OTHERWISE IF they say 4 THEN
d. They count to the third object: 1 . . . 2 . . . 3 . . . 4 from their left.
OTHERWISE IF they say 5 THEN
22 Conjuring with Computation

e. They count to the third object: F . . . I . . . V . . . E from their left.

OTHERWISE IF they say 6 THEN
f. You count to the third object: S . . . I . . . X.
5. Ask them to open the card box and read the message.

Thinking Computationally
The algorithm behind this trick is again a sequence of instructions as
steps 1 to 5 are followed in sequence, but now there is something new,
a new kind of plumbing. In a sequence, every instruction is followed.
Here, there are six possible instructions in step 4 (instructions a–f).
However, only one of them is followed each time that you do the trick.
These instructions are part of a bigger structure that controls the flow:
a selection.
Programmers call this kind of structure an IF statement. It
combines instructions to follow with tests that are either true or false.
For example, “they say 1” is a test. Either they did say 1 (it is true) or
they didn’t (it is false). Programmers call these kinds of test boolean
expressions (boolean means true or false). If the test is true, then
you do the corresponding instruction and ignore the other tests and
instructions in the selection. If the test is false, then you ignore that
instruction and move on and check the next test. Whichever instruction
you ultimately follow, after doing it you ignore the other tests with their
linked instructions and instead move on to the next instruction after the
whole selection statement (here instruction 5). The selection instruction
is just one statement in the sequence of instructions followed in order,
but it controls which of several instructions is actually done at that point
(see Figure 3.2). In fact, you really only need a simple IF statement that
has just a true and a false option. More complicated selection structures
like the one here that choose between lots of options can then be built
from it.
Every time you used a gadget and it gave you a choice, buried
somewhere in the code was a selection statement that a programmer
wrote to do that choice. Do you have a choice of avatars? Do you have
a choice of buttons to press? Or a choice of links to click? Do you have
 Six Magical Objects: Selection 23

Figure 3.2. The flow of control of the IF statement in the Six Magical Objects trick.
It is a sequence of IF statements inside the outer sequence of instructions.
24 Conjuring with Computation

to choose from a menu? What actually makes the machine do something

different in each case is some variation on a selection statement.
Selection is our second kind of control structure. It gives a new
kind of flow of control. To be an algorithmic thinker, you have to
be able to devise and write clear instructions that use both sequencing
and selection.

PAUSE: George Boole

George Boole (1815–1864) was the largely self-taught son of a
shoemaker. He became a great mathematician, inventing a new
kind of algebra, now called Boolean Algebra or Boolean Logic.
It eventually turned out to be ideal for the computer age that
was still almost a century away. Instead of working with numbers
and arithmetic operations, it works with true and false values and
operations like and, or and not. This is exactly what is needed to
describe and reason about the logic circuits from which computers
are built. It is also the basis of how programs make decisions. True
and false values in programs are now called Boolean values as a
result. It was a step forward from the Ancient Greek idea of logic as
originally developed by Aristotle, and so of the idea of logic as the
basis of reasoned argument.
 Chapter 4

Turning the Key

Repetition

Conjuring
A volunteer chooses a card. You unerringly find it even though you
don’t see the card or where it is placed back in the pack.

Computation
The third way of controlling the order instructions are executed
is called repetition. It involves following the same instructions
repeatedly until a test allows you to move on. If you have
sequencing, selection and repetition, you do not need any other
control structure. Other control structures just make instructions
easier to read and write.

The Trick
Have a volunteer shuffle the deck. When they hand it back, secretly
glimpse at the bottom card and remember it. This is your key card.
Spread the deck face down and have them pick out a card without you
seeing it. Tell them to remember it.

25
26 Conjuring with Computation

Figure 4.1. The state of the cards in the middle of the cut.

Reassemble the pack face down on the table. Then ask them to
place their card on the top of this face-down pile. Turn your back, and
ask them to cut the pack wherever they want. Next, have them place the
bottom section onto the top section to complete the cut (see Figure 4.1).
At this point, turn back around. Point out that their card, which you
don’t know, is now buried in the pile at a position that you also don’t
know. This is all true, but unknown to them your key card is now right
next to their chosen card.
Start to deal the cards one by one onto the table. Deal until you
reach your key card, then turn over one more card. Announce that it
was their card.
For a slightly more entertaining variation, you could employ the
“magician in trouble” presentation. Deal three cards past their chosen
card, then pause. Confidently announce “The next card I turn over will
be your card”. Of course, they saw you deal past their card, so they think
you are going to be wrong. Then, in the final twist, you don’t turn over
the next card in the pile you are dealing from. Instead, you pull their
chosen card out of the face-up pile and turn that over instead.
You have saved yourself, with a nice added comic twist.
 Turning the Key: Repetition 27

The Magical Algorithm

Here is the algorithm for the basic presentation.
TO DO Turning the Key:
1. A spectator shuffles the pack.
2. You glance at and remember the bottom card: this is your key
card.
3. You spread the deck face down.
4. The spectator picks out a card and remembers it, but does not
let you see it.
5. You reassemble the deck face down.
6. They put their chosen card face down on top of the deck.
7. You turn your back.
8. The spectator cuts the deck into two piles.
9. They put the bottom pile on top to complete the cut.
10. You turn back around and take back the pack.
11. DO THE FOLLOWING UNTIL you see your key card.
a. Deal the next card face up onto the table.
12. Turn the next card over.
13. Announce that card was their chosen card.

How It Works
Cutting the pack, in the way described, places whatever is at the bottom
of the pack directly before the card that is on top of the pack. Therefore,
as they put their card on the top, you have guaranteed that it will be
the one after the card you remembered.

Thinking Computationally
The third way of connecting instructions to make algorithms is called
repetition. The line
DO THE FOLLOWING UNTIL . . .
indicates we are about to do repetition. We call this kind of instruction
a loop (there are many variations of loops: this is just one kind).
28 Conjuring with Computation

Immediately following those words is a test that is either true or false,

just as in our selection instruction in the last chapter. In our repetition
instruction, either you see the card (the test is true) or you don’t (the
test is false). If the test is true at the point when you check it, you
move on to the next instruction after the loop (here instruction 12). If,
however, the test is false, then you stay in the loop: you do instruction
(11a) and deal with the next card. This is called the body of the loop,
and it is the instruction that is repeated. The body doesn’t have to be
a single instruction. It could be made up of a sequence of instructions
or even contain selection or repetition instructions. The critical thing
about a loop is that after following the instructions in the body, you go
back and do the test again. It may now have changed. Only when you
check it and find that it is true, do you move on to the next instruction.
Otherwise, you keep doing the body and then rechecking the test every
time you finish the body.
To be an algorithmic thinker, you have to be able to create clear
instructions that use all three of sequencing, selection and repetition.
These three kinds of flow of control form the basis of a style of
programming called structured programming.
One of the most important results in computer science is that if you
have a programming language with sequencing, selection and repetition,
then it is as powerful as any other language, in the sense that in
terms of the flow of control, any computation that can be done in
any other language can be done with these instructions. A language
that can do any computation that any other language could do is called
Turing complete. There are many variations of control structures in
languages, but they are not to make them more powerful, and they
are there just to make programs faster, easier to write, or easier to get
right.

Counter-controlled loops
There are other forms of flow of control, but none allow you
to write out any algorithm that you can’t already do with
these three kinds of instruction. They are just variations of these
three.
 Turning the Key: Repetition 29

For the more fun version of the trick, you replace steps 12 and 13
with the new steps:
14. DO THE FOLLOWING 4 TIMES:
a. Deal the next card face up onto the table.
15. Announce that the next card you turn over will be the volunteer’s
chosen card.
16. Turn over the card four places back in the face-up cards.

Here, we use repetition again but in a slightly different form. The first
line tells us we have a loop. Rather than just a test, it says exactly how
many times to repeat the instruction in the body. The line labelled (a)
is, as before, the body, so the instructions are to be repeated.
This kind of loop is called a counter-controlled loop. The difference
here is that we know in advance how many times we need to repeat the
body and just repeat it that many times. We said we don’t need any
new kinds of flow of control though and we don’t. Having a counter-
controlled loop construct like this doesn’t allow you to do anything you
couldn’t do with a general repetition instruction. To do the equivalent
with our original general kind of loop, we would just need to add more
instructions to keep track of a counter. Our test is just that the counter
has counted up to the target number.
Set a counter to 1.
DO THE FOLLOWING UNTIL the counter is 5:
a. Deal the next card face up onto the table.
b. Add 1 to the counter.
All that our new counter-controlled loop structure is doing is packaging
together the general loop and instructions to keep count. It is just a
shorthand for the above. Either way, we had to keep track of how often
we dealt a card. Without the counter-controlled loop, we need to add
the appropriate instructions explicitly. The specialist counter-controlled
loop structure also makes our instructions more readable to a human.
They immediately can see it is a counter-controlled loop from the way
it starts, rather than having to work it out by looking at the whole loop.
Making instructions easy for a human to understand is really important
in programming if mistakes are to be avoided.
30 Conjuring with Computation

Interlude: Penn and Teller

Penn and Teller are an unconventional magical duo, known for
combining magic with comedy. They are masters of the “trick gone
wrong” ending, though their tricks often go wrong in very violent
ways (such as a woman being accidentally sawn in half a second
time, apparently for real). Their tricks are often also framed as being
about revealing how some specialist trick works, only for there to
be a further twist that shows that in their version that could not
possibly have been how it worked.
Penn Jillette has a passion for computer technology and the
web. He was a regular contributor to a computing magazine in the
early 1990s, wrote web articles for a search engine company, and also
appeared in the film, Hackers. Based on this interest, Penn and Teller
invented (and at the time gave away the secret to) a trick called “The
Most Expensive Trick in the World”. It was expensive because of the
then extremely novel, though now every day, advanced computer
science technology it relied upon.
 Chapter 5

Calculator Conjuring
Programming Languages

Conjuring
A friend types a six-digit number into a calculator based on a
number they thought of at random. Miraculously, when you give
them three numbers to divide into it, they all divide exactly. Even
though you had no idea what their original number was, they end
up back with that original number.

Computation
Computer scientists don’t just invent algorithms, they have to be
able to formulate them using the specific instructions available in a
given notation (i.e., a programming language). This involves a
whole new set of skills of being able to work with a fixed notation.

The Trick
Ask a friend to think of a three-digit number. Give them a calculator, or
have them use the calculator on their phone, and ask them to key in the
number. Then, say it would be better to have a bigger number, so have

31
32 Conjuring with Computation

them type in the number again to make a six-digit number. Tell them
that you haven’t seen the number but even so can do some mathematical
magic on it. Point out that the chance of a number dividing by 11,
so giving a whole number answer, is very slim. Suggest they try it.
Amazingly, their number does. Suggest it would be even more surprising
if it was also divided by 7. Suggest they try it. Even more staggering, it
again gives a whole number. What about 13? Surely they couldn’t have
randomly chosen a number that would divide by that too! They do so.
Point out it would be totally amazing if they were now looking at their
original number!

The Magical Algorithm

TO DO Calculator Conjuring:
1. A spectator thinks of a random three-digit number and remembers
it.
2. The spectator keys it into a calculator twice and then presses
ENTER.
3. The spectator divides it by 11 and checks if they get a whole
number.
4. The spectator divides it by 7 and checks if they get a whole
number.
5. The spectator divides it by 13 and checks if they get a whole
number.
6. You point out that magically they have recovered their original
number.

How It Works
The reason this works is all about the number 1,001.
Suppose they thought of the number 123. If you multiply that
number by 1,000 then you turn it into 123,000. If you then add the
original number to this you get 123,123. This works whatever the number
is, as long as it is three digits long. Start with 999: multiply by 1,000
and then add 999, which gives you 999,999. So multiplying a three-digit
 Calculator Conjuring: Programming Languages 33

number by 1,000 and adding the number is the same as making a six-
digit number out of a three-digit number written out twice. Now, this
is also the same as multiplying by 1,001 as the 1 in 1,001 causes the
number to be added to what you get by multiplying it by 1,000. Try it by
multiplying any three-digit number by 1,001. So, when your friend puts
the number into the calculator twice, they are actually just multiplying
by 1,001.
Now, try multiplying 7, 11 and 13 together. The answer is also
1,001. Putting the number end to end and then dividing by 11, then 7
and then 13 is the same as multiplying it by 1,001 and then dividing
by 1,001. Dividing is the opposite of multiplying (mathematicians call it
the inverse), so that is why we get back to the number we started with.
Written out as an equation, we are using the fact:
Secret × 1,001 ÷ 1,001 = Secret
This also explains why it divides by all of 7, 11 and 13. Another way
of writing the number in the calculator is
Secret × 11 × 7 × 13
since 1,001 is just 11 × 7 × 13. This means that all three numbers we
use divide exactly into the total, and when we do each division in turn,
we are just cancelling out that number from the total.

Thinking Computationally
We are mainly using similarities between magic tricks and computing
concepts to help you understand computing. However, here we focus on
a difference that sets them apart.
One thing that distinguishes computer scientists from most magi-
cians is that of precise notation. An important part of thinking
computationally is to be able to write instructions very precisely and
in a precisely defined language: one that can be followed by a computer.
You must be able to get the job done using only the commands available.
Magicians put the onus on secrecy rather than sharing their tricks.
That means that the need for writing down instructions for anyone but
themselves to be able to follow is not a key part of what they do.
34 Conjuring with Computation

Only magicians writing magic books need to worry about the instructions
being easy to follow. Magicians (and likewise cooks) tend to fall back
on human languages like English to write instructions. That is why
sometimes manuals are quite hard to follow as it is quite hard to be
really precise in a human language like English. Using human languages
is usually fine, but they can be ambiguous or just are not precise enough,
with the meaning not always totally clear. If you don’t bake cakes, try
baking one for the first time following a recipe you’ve never seen before
to get the idea. It probably won’t be perfect the first time. The lack of
preciseness in languages is the reason why when we try to follow written
instructions, whether how to do a trick, fix a puncture or bake a cake,
things sometimes go wrong.
Computer scientists, on the other hand, very specifically do want
others (both computers and humans) to be able to follow their
instructions. This means that ultimately algorithms are written out as a
program in a programming language. What that really means is that they
are written out using only specific commands with very precise, agreed
meanings. Learning to program is in part about deeply understanding
what each command does and learning how to convert vaguer plans
into the commands available.
Magicians could use precise languages too, and indeed alchemists,
those pre-chemists-come-magicians who tried to turn lead into gold, did
create notations to write down their “spells”.
A magic language for magic tricks might include commands like
those given in Figure 5.1.
We can implement the algorithm as a program in our language.
It could then be followed by either a human or a robot magician.
This involves not only working out what the specific steps are but also
the detail. Attention to detail really matters in programming. What
exactly do you say and when? What numbers do we allow as a “three-
digit number”. Does 000 count? What about 001? Our program very
specifically rules those out.

1. SAY Magician “Think of a three-digit number”.

2. STORE secret (SECRET volunteer {100 . . . 999}).
3. SAY Magician “Take this calculator”.
 Calculator Conjuring: Programming Languages 35

STORE x n
This means the number n is given the name x in the remainder of the
description of the trick.
SECRET person {n1…n2}
This means the person must choose a number between n1 and n2, keeping
it secret. It delivers a number.
DIVIDE n
This means whoever holds the calculator should divide the current number
displayed on the calculator by n leaving the result displayed on the
calculator and then read out the result. Here n is a whole number.
TYPE n
This means whoever is holding the calculator should type the number n into
the calculator.
PASS person object
This means give person the object.
SAY person s
This means the given person must say s aloud.
In the above, object is a magical prop chosen from {wand, hat, silk handkerchief,
and calculator} and person is chosen from {Volunteer, Magician} n, n1 and n2 are
whole numbers. They could be the name given to a number, an actual number or
an operation (like SECRET) that delivers a number.

Figure 5.1. Part of a language for magicians.

4. PASS Volunteer Calculator.

5. SAY Magician “Type in the number you thought of so it is
displayed on the screen”.
6. TYPE secret.
7. SAY Magician “Let’s make it a longer six-digit number. Type in
the number you thought of again so it is displayed on the screen
twice now. Now I’ve not seen the number so I’ve no idea what it
is. Despite this, I can do some mathematical magic on it.”
8. TYPE secret.
9. SAY Magician “The chance of a number dividing by 11, so giving
a whole number answer, is very slim. Try it to see how lucky you
are!”
10. DIVIDE 11.
11. SAY Magician “Amazing! It does! It would be even more surprising
if it was also divided by 7. Why not try it.”
36 Conjuring with Computation

12. DIVIDE 7.
13. SAY Magician “A whole number. That is staggering. What are the
chances! Let’s try one more. What about 13? Surely you couldn’t
have randomly chosen a number that would divide by that too!
14. DIVIDE 13.
15. SAY Magician “Wow . . . it divides exactly too! . . . And even more
magically, you have ended up back at the number you started
with! That is mathematical magic!”

To be able to write programs like this requires several extra skills over
just thinking up algorithms that computer scientists develop. The first
is the ability to formulate instructions using a specific set of precisely
defined available instructions. That means a programmer needs to be
able to twist what they want to happen to fit what is available. This
is a core skill of programming. They also need the ability to think very
precisely about detail: had you thought about whether 000 was a three-
digit number or not? If you are writing the program, you have to.
Most programming languages are universal meaning that a program
can be written in the language to do any computation that can possibly
be computed. However, the same thing may be done in different ways.
Depending on the problem (and so algorithmic solution developed),
the algorithm as designed might very directly be converted into the
programming language . . . or not. If it can, then once you have the
algorithm, writing the program is relatively easy. In other situations,
the constructs used in your algorithm are far from those available in the
language. We didn’t have a way to say “three-digit number” directly,
for example, so had to work out the range we meant instead. That is
one reason why there are so many languages. Different languages are
designed to make different kinds of problems easier to solve. Choose an
appropriate language for the job at hand, and your task as a programmer
becomes easier.
A key reason we use programming languages is that the instructions
have to be easy to convert into a form that is executable on the
machine. However, there is more to writing programs than that. Just
as importantly, they are for humans to read too; something many
programmers pay too little attention to. Large programs are written
 Calculator Conjuring: Programming Languages 37

by teams, and everyone in the team has to easily understand everyone

else’s code. Furthermore, once written programs have to be maintained,
so changed, both when mistakes are found and when the program is
needed to do something new. That means programming is not just about
writing code that works, but writing it in a way that is as clear as possible
for humans to understand.
In the rest of this book, we will use the slightly more informal
pseudocode notation we used in the early tricks. It mixes English and
notation and aims to be more easily readable. Computer scientists often
use pseudocode notations to help them design both individual algorithms
and code, an important step towards writing the final code. It is less
precise than an executable notation, but easier to work with as you
move from initial ideas of how a program will work towards the precise
detail.

PAUSE: Grace Hopper

Many people have made contributions to the development of
programming languages, but Grace Hopper (1906–1992), ultimately
a Vice-Admiral in the US Navy, made one of the biggest. She
essentially invented the idea of the high-level language that allows
you to write one program that can then be run on lots of different
kinds of computers. Before her, programmers had to write code in
the specific low-level instructions of the machine it was to run on.
Her idea was that programming languages could be based instead
on English, and special programs called compilers could translate
them into instructions for different machines. She also popularised
the word “bug” for programming errors because of a dead moth she
found that was the cause of a computer not working: it was stuck
in the machine.
This page intentionally left blank
 Part 2

Evaluation and Logical Thinking I

To create a correct algorithm, whether one behind a magic trick or one

behind software, you need to be able to think precisely and logically. It
is very easy to get the detail wrong. The flip side to this is that you
must also have ways to check that an algorithm is right. Will it always
do the right thing?
This page intentionally left blank
 Chapter 6

Wizardly Book Magic

Testing

Conjuring
You show that books about magic can control people’s minds. A
volunteer freely chooses a word and then lets the words of the book
jump them randomly from place to place until they land on a final
word. Amazingly, you had placed that word in a sealed envelope,
held by the volunteer throughout.

Computation
It is very easy for programs to contain bugs and subtle mistakes
that mean that sometimes they do the wrong thing. A good
program will work whatever the situation, and to ensure this, we
must actively look for bugs. One way to do this, called testing, is
to run the program many, many times and check it works correctly
every time. With some logical reasoning, we can do better than
that, though.

41
42 Conjuring with Computation

The Trick
First, you need a copy of the book of Shakespeare’s play, Macbeth. You
can buy a cheap copy from any bookshop. (At the end of this chapter,
we will show you how to create your own version of this trick with a
book you already own.)
Write the word “HEATH” on a piece of paper, and place it in an
envelope clipped onto a clipboard. Place on the clipboard, over it, a
piece of paper with the opening lines to Shakespeare’s Macbeth written
out as in Figure 6.1. They are from the famous first scene where the
three witches meet. Draw a dotted line after the first sentence. Write
the last part of the quotation in bold as shown (or write it in red). Make
sure there are no spelling mistakes or missed words.
Now for the trick. Give a volunteer the clipboard and a pen. Give
another person an actual copy of the book, Macbeth.
Tell the audience how later in the play the witches meet Macbeth
and predict that he will be King. But words read aloud over and over gain
a magic of their own. Macbeth has been read so often that the words
of the play can control a person’s mind and so make other predictions
come true.
Ask the person with the book to read the opening lines, while the
first volunteer checks that the words on their clipboard are exactly the
same words that the witches speak.

When shall we three meet again

In thunder, lightning, or in rain?
-----------------------
When the hurlyburly’s done,
When the battle’s lost and won.
That will be ere the set of sun.
Where the place?
-----------------------
Upon the heath.
There to meet with Macbeth.
I come, Graymalkin!

Figure 6.1. The first lines of Macbeth written out as needed for the trick.
 Wizardly Book Magic: Testing 43

Get the volunteer to put their hand on the book and “draw on its
magical powers”. Have them then pick a word at random from the first
sentence of the quotation (before the first dotted line), as spoken by the
first witch. It should be one of the words:
When shall we three meet again/In thunder lightning or in rain.
They should circle their chosen word, telling everyone what it is. For
example, if they choose the word “three”, then that is the word they
circle and announce.
Now, explain that they must let the book control them. They
will jump from word to word, guided only by the magic of the words
themselves. They will stop only when they end a jump on a bolded
word (after the second dashed line). That bolded word will be their final
chosen word.
To do this, they should count the number of letters in the original
chosen word. They then step forward that many words to land on a new
word. They circle it. For example, if they did choose the word, “three”
originally, then it has five letters. They step forward five words and land
next on “lightning” which they circle. Have them do this with their word.
Double-check that they don’t make a mistake.
They do this repeatedly bouncing from word to word, each time
counting the number of letters in the new word and moving on that
far. They should ignore punctuation. They should also ignore the stage
direction in the book which is why you’ve given them a version with only
the words spoken.
They keep going until they jump to a bolded word for the first time.
Check each step with them and the audience. They should announce
what that final bolded word is. Have them agree that it was their own
free choice of which word to choose originally and that they had no idea
what bolded word they would land on. No one knew.
Finally, explain that using the magic in the book, you made a
prediction beforehand. Point out that there is an envelope on their
clipboard under the sheet they have been writing on. They should open
it and read out what it says. Magically, they find the word written is the
bolded word that they finally landed on!
44 Conjuring with Computation

The Magical Algorithm

TO DO Wizardly Book Magic:
Before the trick starts
1. Write HEATH on a piece of paper.
2. Place it in an envelope.
3. Place the envelope on a clipboard.
4. Cover it with a piece of paper with the words to Macbeth on
exactly as shown.
The trick itself
5. One volunteer reads the words from the book, while a second
who is holding the clipboard checks they are the same.
6. The volunteer with the clipboard chooses and circles a word from
the first sentence.
7. DO THE FOLLOWING UNTIL the last word circled is a bolded
word.
a. The volunteer counts how many letters are in the last circled
word.
b. The volunteer jumps forward that many words.
c. The volunteer circles the new word.
8. The volunteer tells everyone that chosen bolded word.
9. Tell them about the prediction in the envelope.
10. The volunteer reveals the predicted word in the envelope.

How It Works
Whatever word they pick from the first sentence, your prediction will be
correct (of course). Each word in that first sentence leads to the same
bolded word. We will come back to exactly why it works after we have
looked at some computational thinking, though.

Deeper magic?
Now you know the trick, you can do it yourself, even if you don’t
understand how it works (yet). If you start with the words from Macbeth,
it will always work. In fact, it is even more magical. It will work with just
about any book about witches, wizards or magic, though the word to
 Wizardly Book Magic: Testing 45

predict will be different for each. Try it with the words from the start of
a different book: The Wonderful Wizard of Oz. Don’t trust me to give
you the words. Find a copy. For this book, if you pick a word from the
first sentence, you will always end up on the word “FOUR”.
Want more evidence? Try another completely different magical
book: The Cat in the Hat by Dr Seuss. It’s about a magical cat. Choose
a word from the first sentence. I predict you will land on the word “SAT”.
The magic is even deeper than that. It will work with just about any
writing, though in some books the sentence to make bold is further on
than in others.

A conspiracy?
Is this really some powerful, deep magic? Or has there been a secret
conspiracy by authors, all agreeing for centuries to write their books
so this trick works? No. It’s just a simple, if surprising, property of the
words in books. It is another “self-working trick”, another algorithm. It
always works if you follow the steps, whatever book you choose.

Thinking Computationally
Testing times
I’ve claimed the trick always works for that passage from Macbeth. Do
you believe me enough to try the trick for yourself with an audience?
How can you be sure it always works? Well, you could try it a few times.
Would that convince you? Does that give you enough evidence? What
if they pick a word you didn’t check? You have no idea whether it works
for that one. You will look a bit silly if your prediction turns out to
be wrong.
It sounds a bit tedious, but you should check every alternative word
that could be chosen does work. You have to be very patient to be a good
magician or a good computer scientist! Careful checking and attention
to detail really, really matter. Miss a step and things can go wrong.
Programmers call checking programs like this, testing. Programmers
spend more time testing programs than writing them. Testing is one
kind of evaluation of a program: evaluating if it works.
46 Conjuring with Computation

We want to be sure we really have checked all the possibilities.

The easiest way to do this for our trick is to start at the first word,
check it works, then move to the next, check it, and so on, until we
have done all words in the first sentence. That means for Macbeth
you need to do 12 tests as there are 12 possible starting words. Other
books will need more or less tests depending on the length of the first
sentence.
We are doing algorithmic thinking again: what I just described is
an algorithm for checking our magic trick works. Since we now have
an algorithm to do it, we could even write a program based on it to
check our trick rather than do it by hand. If you can write programs,
you might want to try. If not, check the magic trick really does always
work for Macbeth, The Wonderful Wizard of Oz and The Cat in the
Hat, by hand.

The lazy logical thinker does less

With a little thought, we can actually come up with a slightly better
algorithm that involves less work to do the checking. Here is how. Circle
each word you come to as you check the paths of each word. Then, if
you land on a circled word from an earlier test, stop and move on to a
new starting word. You know the path from that point on has already
been checked.
Figure 6.2 shows the result of checking every word from the first
sentence in Macbeth like this. We have added arrows showing where
each word jumps to so that we can see the paths taken. Remember, it
is only the paths from words in the first sentence you need to check as
those words are the only ones that can be chosen.
As part of checking the word “when”, for example, we bounce to
“meet” and then to “lightning”. When it is the turn of those words to
be checked, we will have nothing more to do as having checked the trick
works for “when” we have already checked those other words too: they
follow the same path.
With a bit of logical thinking, we have reduced the number of paths
we have to check from 12 to only 7, almost halving the work to do.
Some of those paths are shorter, so quicker too. Perhaps you can see a
way to reduce the number of paths that we need to check even further?
 Wizardly Book Magic: Testing 47

Figure 6.2. The paths followed in the book Macbeth.

You have to be very careful when checking. It is easy to make a

mistake counting. If you do, then you might find one day your trick
doesn’t work after all. That means it’s best, having done it once, to
double-check your work. Drawing the arrows makes that slightly easier.
You just need to check every word has an arrow out, and every arrow
jumps the right amount.

Why does it work?

Why does it work? Think about what happens in that first sentence as
the paths of some words, like “when”, “meet” and “lightning”, join up.
Once paths have joined, they don’t split up again. Any short word in
that first sentence is just going to join with another word that could
have been chosen. They will end up on the same bolded word. Only a
few words, mainly ones at the end of the sentence or very long ones,
will jump straight to the second sentence. In fact, only four words jump
48 Conjuring with Computation

out of that first sentence: “thunder”, “lightning”, “in” and “rain”. What
appeared to be 12 possibilities for the volunteer to choose from is actually
only four different choices of onward paths: only four paths cross the
first dotted line. A similar thing will happen with any book. Long first
sentences may look like they are offering lots more choices but actually
only a few words will jump the first dotted line and escape as separate
paths from that first sentence. Once we are beyond the dotted line, no
new paths are created. We don’t have to check every word from that
point, only those we can land on from one of the existing paths.
So what happens to those four paths for Macbeth? Does that still
mean there are four possible bolded words we could end up on? Well, no.
As their paths bounce around, eventually by chance two paths hit the
same word. Then those two paths merge and stay together from then
on. Once they do, there is one less possibility for what the final word
could be. That is what happens in the diagram for Macbeth. The further
we go, the more likely it is that the different paths left will eventually
meet and so land on the same bolded word. For Macbeth, they all meet
31 words from the start. If you go far enough, it is virtually certain all
paths will join up whatever the book. You just may have to push the
bolded sentence back.

An algorithm to create your own book trick

To create a version of the trick for your favourite book, copy out the
start of the book. Draw the start line after the first sentence. Then,
check all the paths from words of that first sentence. Find the earliest
word they all meet on (if they do). The finish line is the end of that
sentence. Make the next sentence the bolded one. Your predicted word
is the word where the now single path jumps to first in that bolded
sentence. To be extra safe, double-check that every word in the first
sentence does take you there!

Bugs
Programmers call the mistakes they make bugs, and finding mistakes is
called debugging. So when double-checking your version of the trick, if
you found a word it didn’t work for (and so pushed the bolded sentence
 Wizardly Book Magic: Testing 49

back to fix it), you are essentially debugging your version of the trick.
The word bug had been used before its use in programs as a mistake but
caught on in programming supposedly because one of the first times a
computer went wrong, it turned out it wasn’t because of a programmer’s
mistake, it was because a real bug (a moth) got trapped in one of the
switches! One of the first programmers, Grace Hopper, stuck the moth
in her notebook and the name stuck too.

All books?
Does it work for all possible books? How could we check that? We would
have to work through every book in turn and see if the paths for each did
join together as in Macbeth. If ever we found one that didn’t work, we
would have proof in that book that it does NOT always work. However,
each time we check a new book that it does work for, we’ve proved
nothing for certain about all books. We would have to keep testing to
be sure it works for the next book, and the next. . . . To prove that it
worked for every book (if it does), we would need to check every single
book ever written. That would take forever! Let’s hope it doesn’t always
work! We need some logical thinking.
If it doesn’t always work, we should be able to think of a way that a
book might be written so that at least two paths don’t join up. Well, if
two words are the same length, then they will jump to positions the same
distance apart. If that keeps happening, their paths will never meet. So if
we have a book that is something like the following all the way through,
then the trick won’t work.

The dog saw the rat,

but the cat ate him.
The dog was sad for the rat. . .

Here, all the words are three letters long, so we end up with only three
paths straight away. However, those paths just bounce along together,
never joining up. Check it. If every word in a book is the same length
like this, the trick will fail to converge on a single word however far you
go. So without checking any real book at all, we have worked out that
50 Conjuring with Computation

it is possible for the trick not to work on some books. (Though we still
don’t know if anyone ever wrote a book like that.)
With similar reasoning, we can also see what will help make it work
quickly. The paths will generally join quickly, for example, if the book
uses lots of short words but that have different lengths: say lengths 2, 3
and 4 letters.

Avoiding disasters
Just as we originally tested the trick for all possibilities to prove it works,
programmers test their programs over and over again to make sure they
work, whatever happens.
When a computer program, perhaps a game you are playing, crashes,
it means you have just found a situation that the programmers didn’t
test. It isn’t your fault, it’s theirs. It is virtually impossible to write
complex software, without it containing mistakes somewhere. Why?
As real programs can be millions of instructions long, it takes perfect
attention to detail and logical thinking to get it right. It is just too easy
for the programmer to make a mistake, for a program to contain a bug.
Testers don’t find all the problems because there are just too many
possibilities to check them all. Testing real software is actually more like
checking the trick works for all the books in the world than checking
that Macbeth or The Wonderful Wizard of Oz works. There might be
millions, billions or more possibilities to check, and it just can’t be done
in any sensible time. Instead, programmers often test what they hope is
a good enough sample. They then cross their fingers and hope. That is
why programs are constantly being updated with “bug fixes”. It is only
when millions of people start using the software that it starts to get
properly tested and the bugs are found. They are using their customers
as their testing team.
Testing is an important part of any evaluation, and the testers have
to test as completely as possible. But testing alone just isn’t good enough
to find the problems in really complex programs. There are just too many
possibilities to check. There are also problems testing won’t find.
Better ways are needed.
 Wizardly Book Magic: Testing 51

Especially, for critical systems, programmers use other evaluation

methods too. As we did for our trick, they use logical thinking to work
out what to test, trying to ensure they cover, if not every situation, a
full range of situations.
Both magicians and programmers must check that their algorithms
always work, whether by testing, logical reasoning or both. This requires
a lot of patience, care and attention to detail.

PAUSE: Margaret Hamilton

Testing is a key part of software engineering, which is the idea that
programs are not just written but engineered. Margaret Hamilton is
a key person involved in the history of how programming was turned
from a craft into such an engineering discipline. She was part of, and
ultimately led, the team who developed the code for the command
module of the Apollo moon landings. Hamilton is responsible for
the idea of end-to-end testing where the full system is tested in
realistic conditions. One of her other innovations was the priority
alarm display, which was included to deal with the problem of the
computer becoming overloaded and not being able to do all that
was asked of it. When the display showed warnings, Buzz Aldrin
was supposedly told to count to 5 before hitting any button to allow
it to catch up. Hamilton’s foresight actually did help avoid the moon
landing being aborted 3 minutes before it was due to touch down.
This page intentionally left blank
 Chapter 7

The 21-Card Trick

Logical Reasoning and Proof

Conjuring
A volunteer picks a card and thinks about nothing but that card
throughout the trick. You show you can read their mind by telling
them their card, even though they told no one of their choice.

Computation
Real programs are too complex to test all possibilities in any
reasonable amount of time. A complementary method is to use
logical reasoning to give a convincing argument or proof as to why
a program or specific algorithm always works without doing any
testing. By doing both, you increase confidence that the program
really is correct.

The Trick
Tell a volunteer that they must think of a card. You are going to read
their mind and work out which it is. Explain they must think hard about
the same card all the way through the whole trick. First, you need some
cards for them to choose from.

53
54 Conjuring with Computation

Shuffle a pack of cards and then deal 21 cards face up into three
piles of seven cards. Deal one card at a time to each row, before moving
back to the first pile and so on, making sure all cards are visible.
Ask them to secretly pick any one of the 21 cards and think of it
and of nothing else. Explain that if they think of anything else or giggle
or laugh, it will make it really hard to read their mind. Go up to them
and stare at their forehead, concentrating, warning them not to giggle.
If they do giggle or smirk, say that spoiled it. Either way, say they seem
to be thinking of other things so it’s making it hard. You will need to
try again. Ask them to point to the pile their card is in, without telling
you which card it was. You collect the cards up and deal them out again
for another attempt.
When you collect the cards, collect them in their three piles, placing
the pile the volunteer pointed to in the middle. Then deal out the cards
again in the same way into three equal piles, across the rows.
Now as you couldn’t read their thoughts through the front, the front
of the skull is after all pretty solid, try the back this time. Remind them
not to move and especially not to giggle but just think of their card.
While they do, stand behind them and concentrate on the back of their
head.
Again, explain something is going wrong: perhaps they moved or
giggled or someone else made a noise. Say you almost had it: you think
you got the colour this time. You will need to try again.
Once again, ask them to find their chosen card and only tell you
which pile it is in. Collect the cards in the same way as before, casually
placing the new pile containing the chosen card in between the other
two piles and dealing them out a third time.
As their skull is obviously hard for thoughts to go through this time,
you will try through the ear. Again, they think of the same card and you
stare at their ear from one side. This time, announce you think you’ve
got it: it is so much easier to read the thoughts coming out of a person’s
ears, but say you want to triangulate and check from a different angle.
Have them point to the pile their card is in, collect the cards in the
same way with their pile going in the middle and deal them out again
across the rows for a final time.
 The 21-Card Trick: Logical Reasoning and Proof 55

Now, go to their other side and stare at their other ear. Announce
that yes you’ve got it and go to the pile of cards. Stare at them saying
you’ve now just got to find it. Turn over all of the first pile saying it
is not one of them, checking with the volunteer that you are right so
far. Next, turn over the last pile: it is not one of them. That leaves
the middle pile. Start to turn most of them over, then pause, go back
and turn over the bottom three. Say that it is not any of those, then
quickly turn over the top three too, leaving the single middle card face
up. Announce that that is their card, asking them to confirm you are
right.

The Magical Algorithm

TO DO the 21-Card Trick:
1. Shuffle the pack of cards.
2. Deal the first 21 cards one at a time to make three piles of seven
cards with all cards visible.
3. Ask a volunteer to secretly pick a card from the 21 and remem-
ber it.
4. DO THE FOLLOWING 3 TIMES:
a. Ask the volunteer to think of their card and nothing else.
b. Pretend to read the volunteer’s mind but say you haven’t quite
got it yet.
c. Ask them to point to the pile their chosen card was in.
d. Pick the piles up sandwiching the one pointed to between the
other two.
e. Deal out the cards again, dealing the cards one at a time across
the piles.
5. Deal out the cards a final time.
6. Ask the volunteer to continue to think of their card and nothing
else.
7. Pretend to read the volunteer’s mind.
8. Reveal that their chosen card is the one in the centre of the middle
pile.
56 Conjuring with Computation

How It Works
There are two important steps to this trick. You must always deal out
the cards across the piles. You must also always take the pile they point
to and put it in the middle of the other two piles when you pick up the
piles. Do that and their card is guaranteed to end up as the middle card
of the middle pile after the fourth deal.
Before you read on, have a go at working out why that works.
Perhaps try it a few times. Clearly, no one can read your mind. We
are just following the steps: the same ones every time. It seems to never
fail (as long as the volunteer keeps thinking of the same card all the
way through and does not lie about the pile it is in, of course). To trust
the magic enough to do it for a live audience, we must be pretty sure it
always works, though.

Thinking Computationally
One way you could convince yourself is by doing the trick multiple times:
by testing. Would that convince you? Try it. What about the next time
you do the trick? Will the rules it is following work that time too? How
many tests are enough?
You could test every possible way the trick could start and see
how it ends. That’s what we did first with the Wizardly Book Magic
trick. In that trick, it was a bit boring, but it was possible, as there
were only 12 different situations to check. What are the possibilities
here? There are 52 cards in a pack and any 21 could be used in any
order. We would have to test all those different possibilities. There are
9,809,042,663,139,505,407,817,600,204,800,000 permutations to try.
That’s a lot of times to do a magic trick. For each, we would need
to test the case where the volunteer chose each of the 21 cards. So do
that 21 times over to be completely exhaustive. That is a lot of testing.
You would have to stop eventually. Or is there a better way?
Well, as with the book trick, we can do some simple logical
reasoning. Here, it is vital. Maybe you worked this part out already. We
don’t actually have to do all those permutations as it’s not what’s on
the card that actually matters to the trick. It is only where the chosen
 The 21-Card Trick: Logical Reasoning and Proof 57

card is at the start that matters. If, from each starting position, the
chosen card always ends in the middle, the trick always works whatever
the cards actually are. Thinking that through, it means there are actually
only 21 tests we need to do, one for each place the chosen card could
start from. It is actually possible to do the trick 21 times to check what
happens in each case (if you’ve an odd hour to spare and are meticulous
about keeping track of what positions you check).

Prove it!
However, we can do better than that. In fact, I know it always works
without any tests. I’ve proved it (so I can sleep easy at night now before
a magic show). With a little logic and proof, I can be sure the rules are
right for all 21 possibilities without testing any.
This magic trick works because putting the pile containing the
chosen card in the middle of the other two piles and re-dealing the
cards in effect limit where their card can go. Let’s step through it.
After Deal Number 1: After the first deal of the cards into three piles,
the chosen card could be anywhere, but when we pick up the pile pointed
to, it goes into the middle seven-card pile. Why? Because we put that
pile (whichever it is) in the middle of the other two. There are now
only seven positions it could be in within the pack (positions 8–14 in
Figure 7.1).
After Deal Number 2: You deal the cards into three new piles dealing
across the piles. Where do those seven cards (numbered 8–14) from
the selected pile go from the middle pile? They end up in the middle
positions of each pile (see Figure 7.2). The chosen card has to now be
in one of those seven places. Those places are the fourth or fifth card of
the first pile; the third, the fourth or fifth card of the middle pile, or the
third or fourth card of the last pile. You ask the volunteer to say which
pile, and again put that pile between the other two. Wherever the seven
cards were, they are moved to the middle pile. They now must be one
of the middle three cards in the middle of the pack (as shown by the
arrows of Figure 7.2).
After Deal Number 3: After this deal, the chosen card has to be the
middle card of the first, middle or last pile (see Figure 7.3). Why? It was
58 Conjuring with Computation

Figure 7.1. The positions of the cards after the correct pile is pointed to and moved
to the middle the first time. The cards are numbered by positions 1–21 at this point.
Here, we draw the piles as if still laid out. In the trick, they would be in this order
in a single pile at this point. The selected card will be ultimately in one of the
positions 8–14 in the collected deck.

Figure 7.2. The positions of the cards after the second deal. A similar pattern follows
if it was in one of the other piles. If the chosen pile is not in the centre, then when
the piles are collected up they will have been moved to one of the positions: 8, 11
or 14.
 The 21-Card Trick: Logical Reasoning and Proof 59

in one of those middle three cards in the middle pile. By dealing across
the rows, they all get dealt one after another, one to each pile. Once
you are told which of the three piles has the card this time, you already
know exactly where their card is. However, when you collect the piles,
whichever of the three it is, it is moved to the middle (as shown by the
arrows in Figure 7.3).
After Deal Number 4: The fourth deal simply moves the chosen card
to the middle card of the middle pile (see Figure 7.4), which really is
just for effect.
What we have just done is give a convincing (we hope) argument
that the trick, or actually the underlying algorithm, always works. That
is all that mathematical proofs are: convincing arguments where there
is no room for doubt if you follow the detail. Here, we were just proving
that a trick works, but in doing that we have shown we can prove
algorithms are correct, more generally. Now, computer programs are
based on algorithms. So we can do the same thing for a program and
prove it works in a similar way. It is very important that programs always
work too.

Testing versus proof

The normal way programmers make sure that a program works is to
test it: that is, try it out lots of times. Testing every combination for
a program is even more impractical than for our trick. There are many
more possibilities to try for any realistic program. Instead, as many as
possible are tested. If it works each time, then you assume it also works
in the cases you didn’t try (and hope!). To improve on this, as we did,
testers use logical reasoning to cut down the number of possibilities to
test to be more sure it works. Still, for real programs, this leaves more
tests than is possible to be totally sure.
The sort of proof we did gives an alternative. If the proof has
no flaws, then it proves the trick (or program) works whatever the
combination is, and you don’t strictly need to test any of them. In
practice, it is a good idea to do lots of testing too, as proofs can have
mistakes in them just like programs. However, if you want to increase
your certainty that your trick (or program) works, try and prove it!
60 Conjuring with Computation

Figure 7.3. The positions of the cards after the third deal. On moving the selected
pile to the middle, the selected card will have ended up at position 11.

Figure 7.4. The positions of the cards after the fourth deal. The chosen card that
was collected into the middle position stays there as the cards are dealt across the
rows. It ends in position 11.
 The 21-Card Trick: Logical Reasoning and Proof 61

Proofs and new versions of tricks

The beauty of doing proofs is they can give you a really deep
understanding of why a trick works. Once you understand the mechanics,
you can play with changing some of the presentation. As long as you
always make sure it doesn’t affect the proof, or you can adapt the proof
along with the trick, it will still work.
With this trick, we can see from the proof, and so understanding
why it works, that the order of the cards in the chosen pile must not be
changed. However, the two other piles could each be separately shuffled
before being put together. Adding that step doesn’t affect the proof. As
long as the chosen pile goes undisturbed between the two other piles
of seven cards, the order of the other cards doesn’t matter. The proof
allows us to see what will stop the trick working and what won’t. Perhaps
you can come up with your own twists now you know how it’s done.

Phone crashes, rocket crashes

For games, bugs are just irritating, but there are many situations where
they can cause really big problems. For example, when you pick up
a phone to dial the emergency services, you don’t want to find the
phone is dead because of a software bug. On January 15th, 1990, it
happened across the USA: 70 million telephone calls failed. Telephone
company AT&T’s engineers had upgraded the program that makes the
connections, so each phone links to the one it is calling. The code in all
their 114 switching centres was changed. The new version had a bug,
and they all stopped working properly. The mistake was in a few lines
of code out of millions, and the programmers who had changed them
hadn’t tested all eventualities. AT&T lost $1 billion as customers fled
to competitors.
NASA’s Mariner 1 spacecraft was launched in 1962. It was to fly by
Venus. It made it as far as the Atlantic Ocean. A hyphen that should
have been there was missed in the program: a simple error, but lethal in a
computer language. Testing didn’t find the problem. The flight program
miscalculated the rocket’s trajectory, and the rocket lost control. It had
to self-destruct before it caused bigger problems. The mission lasted 290
seconds.
The Ariane 5 rocket had taken 10 years to build and cost $500
million. In 1996, it exploded 40 seconds after liftoff, destroying its
62 Conjuring with Computation

satellite cargo. This spectacular software failure was due to squeezing

a large number into the computer memory reserved for a small one. As
it went faster and the number got larger, there wasn’t enough space to
hold the rocket’s speed. This caused the rocket to veer off course, break
up and explode. The testers had missed testing this critical situation, so
no one realised the mistake until it was too late, though 40 seconds into
a real flight was long enough to find it!
Bug-caused disasters like these show why we need as many ways
as possible to avoid bugs being missed in mission-critical software.
Arguably, all software should be evaluated more thoroughly. Make sure
you do all you can to ensure your tricks always work if you want to avoid
onstage disasters.

PAUSE: Edsger W. Dijkstra

Dutch Computer Scientist, Edsger Dijkstra (1930–2002), amongst
many amazing contributions to early computer science, revolu-
tionised the way programmers thought about writing programs
from the 1960s onwards. He introduced the idea of Structured
Programming which led to changes in the design of programming
languages. It is a style of programming that helps improve the
quality of programs making them easier to write correctly. It involves
systematically using specific control structures like selection and
repetition, avoiding “spaghetti code” where control is based on jump
or goto statements. This makes code both easier to get right and
easier to reason about.
 Chapter 8

Mental Monte and the Trains

of Thought
Reasoning by Cases

Conjuring
A volunteer picks a card from three, then moves cards around, all
with your back turned. After more mixing of the three cards, you
can still name the card chosen.

Computation
One form of logical reasoning is reasoning by cases. This involves a
form of decomposition: splitting your argument into several cases
that together prove the whole. It is naturally used at decision points
in a program, so with selection instructions. There is one case for
each path through the code.

The Trick
Three-card Monte, or Find the Lady, is a popular game which is
often used to swindle people. Three cards are used, one a red queen.

63
64 Conjuring with Computation

The cards are moved around, and the spectator tries to find the queen,
normally unsuccessfully. Whilst this game can be rigged by sleight of
hand, we present a variation, Mental Monte, that makes it more magic
than scam.
Put three cards face up on the table and ask your volunteer to choose
one mentally and remember its position. You turn your back and ask the
volunteer to point at their card, to show the audience which card they
chose. Then have them silently switch the two other cards, the cards
not chosen. They then turn the three cards face down. When you turn
round, you ask the spectator to switch pairs of cards in the row of three
as many times as they wish, so mixing up the already mixed-up cards.
You then turn the cards over. Even after this multiple mixing, you are
able to read the volunteer’s mind and reveal their chosen card.
Now, you have the basic idea, how about an alternative presenta-
tion? You can perform this trick with three train tickets for three different
destinations. If you don’t have the tickets, just write three places on three
bits of identical cards. In this presentation, you can talk about changing
the destination, and about long complex train journeys. This gives a
natural justification for moving the face-down tickets around. It also
gives a strong finish when you reveal the name of the mystery-selected
journey.

How It Works
The key to this trick is that you memorise the card in the middle of the
three when they are face up at the start. When you turn back around,
you need to casually keep an eye on the position of the face-down card
in the middle, and track it as it is moved around in the swaps. Once the
swapping is finished, you flip over the card you have been tracking.
There are a few possibilities. If the card you flip over is the card you
had originally memorised, then you correctly announce it as the chosen
card. If it is not that card, then the chosen card is neither this card nor
the one you memorised. Flip over the others, and announce that it is
the other card (the one neither tracked nor memorised).
To understand why this works, we will need some logical thinking . . .
 Mental Monte and the Trains of Thought: Reasoning by Cases 65

The Magical Algorithm

TO DO the Mental Monte:
1. Place three different cards face up on the table.
2. Memorise what the middle card is.
3. A volunteer chooses one card and remembers its position.
4. Turn the three cards face down.
5. Turn your back.
6. The volunteer points at their card to show the audience.
7. The volunteer switches the two unchosen cards.
8. The volunteer turns all cards face down.
9. Turn back around.
10. The volunteer repeatedly switches pairs of cards while you
secretly keep track of where the middle card ends up.
11. Flip the tracked card over.
12. IF the card you tracked (so flipped over) is the memorised card
THEN
a. announce the flipped one is the card they chose.
OTHERWISE
b. turn over the other cards and announce the third card (not
the one you memorised or the one you turned over first) is
the card they chose.

Thinking Computationally
To use logical reasoning to be sure this trick always works, we need to
work through the separate cases that can occur. The algorithm has a
selection statement, that means you do something different depending
on what the card you turn over is. That IF statement gives us two cases
to consider. The case when the card flipped over first is the memorised
card, and the case when it isn’t. We must therefore think through those
two cases.
66 Conjuring with Computation

The THEN case

If the card you flip is the one memorised so is still the one tracked, then
logic tells us the spectator didn’t switch the middle card when your back
was turned. If they had moved it, it would no longer be in the middle
when you turned around so not be the one you then followed. They must
have swapped the two end cards. However, they were told to swap the
cards they didn’t choose, so the original middle card was their choice.

The OTHERWISE case

If the card you tracked from the middle and so flip is not the one you
memorised, then that card must have been swapped from the middle
position by the volunteer while your back was turned. If it had not been
swapped, it would still have been in the middle and so be the one you
tracked. As the non-chosen cards were swapped, the card you flipped
over was not the chosen card. The chosen card must be one of the other
two cards.
Now the memorised card was originally in the middle. It must have
been swapped from the middle as we just saw. Therefore, it also cannot
have been the chosen card, or else it would have been left in the middle
while your back was turned. It would then have been tracked and be the
one you turned over. It wasn’t so must therefore have been moved in
the original swap.
That means the chosen card, the one that wasn’t swapped while
your back was turned, must be the other card: neither the memorised
nor the tracked one.
Did that just make your head hurt? It may help to understand this
second situation better by thinking through an explicit example. Let’s
suppose the cards at the start are cards 1, 2 and 3, in that order. The
middle card is card 2, and you memorised it. Suppose, when your back
is turned, the spectator chooses card 3. That means they switch cards
1 and 2 giving the new order 2, 1 and 3. See Figure 8.1.
You turn back around and follow the middle card, card 1. You track
its position as the subsequent pairs of cards are swapped. You turn over
card 1 wherever it ends up, as it’s the card you have been tracking. It
is not the card you memorised (2), so they must have been swapped so
 Mental Monte and the Trains of Thought: Reasoning by Cases 67

Figure 8.1. If the third card is chosen, the other two are swapped, one being the
memorised one, the other the tracked one.

neither is the chosen one. The card they picked was therefore neither the
card you memorised (2) nor the one you just turned over (1). It must
be card 3.
As the situation is symmetrical, the same reasoning applies if they
instead chose card 1. This time they swap 2 and 3 giving the order 1,
3 and 2 and because it is in the middle when you turn back, you track
and eventually turn card 3 over. Again the card they picked was neither
the memorised card (2) nor the one turned over (3). It must be card 1
(see Figure 8.2).
Our overall logical reasoning is a proof by cases, with one case for
each possibility resulting from the IF statement. This idea is the key to
logical thinking around algorithms that use selection. You must consider
68 Conjuring with Computation

Figure 8.2. If the first card is chosen, the other two are swapped, one being the
memorised one, the other the tracked one.

each of the cases. The structure of the algorithm describes the structure
of the logical thinking you must do.

Interlude: Bob Hummer

The Mental Monte trick was devised by the great Bob Hummer
(1906–1981), a creative magician who performed in and around
Chicago. He invented lots of tricks including lots with mathematical
angles. Mathematician, Martin Gardner, included some of these in
his books on mathematical diversions.
 Part 3

Making It Work for People

To develop software that is easy to use, you need to understand people,

and in particular the limitations of our brains. Great computer scientists
need to understand cognitive psychology.
This page intentionally left blank
 Chapter 9

The Cyclops' Eye

You Cannot Trust Your Eyes

Conjuring
Stare at the picture of the Cyclops’ eye, and it starts to magically
move. Don’t look for too long or you will be forever trapped by its
stare!

Computation
When writing software, you have to understand that our brains
have limitations that lead us to make mistakes. You must write
the software (and especially its interface) in a way that works with
those limitations. If you ignore our limitations, you will produce
bad software.

The Trick
This trick is just a few steps: warn everyone that they should not look
for too long or they will be trapped forever by the Cyclops’ stare, have
everyone look at the picture of the Cyclops’ eye (Figure 9.1) and move

71
72 Conjuring with Computation

Figure 9.1. The Cyclops’ eye was created by Kelly Burrows based on an illusion
invented by Japanese artist, Hajime Ōuchi, in his book Japanese Optical and
Geometrical Art.

their head slowly from side to side. Everyone who looks at it will see the
central part of the eye start to hover and float around.

The Magical Algorithm

TO DO the Cyclops’ Eye:
1. Warn everyone that they should not look for too long or they will
be trapped forever by the Cyclops’ stare.
2. The audience look at the image of the eye.
3. The audience move their heads gently from side to side.

How It Works
As with other optical illusions, this is just your brain playing tricks on
you. Your brain is fooled because it separately tries to work out what is
going on with the horizontal and vertical patterns. As your eyes move,
the strong patterns seem to be doing different things, so your brain
makes sense of this by assuming they are moving separately.
 The Cyclops’ Eye: You Cannot Trust Your Eyes 73

There is another illusion, here too, although it is one that we perhaps

take more for granted. We see the outer shape of the white of the eye
even though there is no actual line. Our brains fill in the shape of the
eye as though there was a shape outlined there.
Optical illusions like this occur because, in order to process the
vast amount of data coming into our eyes, our brains make simplifying
assumptions. We have some hard-coded rules that we cannot switch off
that illusions are designed to trigger. Essentially, this shows we do not
see the world as it really is. It is as though we are living in the Matrix
of the film, seeing a made-up version of the world. However, it is not
some alien machine that is feeding us a model of the world we see, but
our own subconscious brains.

Thinking Computationally
This is our first indication that to be good at developing usable software
you need to be much more than a logical, algorithmic thinker. You must
understand the way our brains work, and in particular their limitations.
As we will see, some of the limitations and simplifications our brains
make mean we are highly likely to make mistakes using poorly designed
software. When creating software, you must take these limitations into
account when you design the user interfaces of software (the parts
that people see and interact with). Good computational thinkers need to
understand people too. Most software now uses some sort of graphical
user interface: one that tries to make tasks easier to do based on the
limitations of our brains. They can be well designed, so easy to use, or
poorly designed, so much harder, with mistakes likely.
74 Conjuring with Computation

Interlude: Hajime Öuchi and

Bridget Louise Riley
The central illusion in the Cyclops’ Eye was published in 1977 in
a book called Japanese Optical and Geometrical Art by Japanese
artist, Hajime Ōuchi. The book was intended as a sourcebook of
ideas for illustrators. The illusion was celebrated and investigated by
visual perception researchers, though it was almost 40 years before
they tracked down its creator. It follows a tradition of Op Art (optical
illusion-based art). The artist best known for pioneering this kind of
art was Bridget Riley. She became famous in the 1960s for her black
and white geometrical work which gave the illusion of shapes and
movement. She had been inspired to experiment with art based on
optical illusions by the pointillism paintings of Seurat that used small
pixel-like circles of colour.
 Chapter 10

Magically Weighted Boxes

Your Brain Cannot be Trusted at All

Conjuring
At the snap of your fingers, you make boxes clearly heavier than
they were only a moment earlier.

Computation
Our brain’s limitations are not just about what we see, but about
our other senses too. Multimodal software that makes use of our
other senses can make the software easier to use, not least for
those with disabilities. However, the software developer must take
into account the limitations of the senses being used.

The Trick
Beforehand, take three empty playing card boxes. Stuff the middle one
with coins to make it heavy and tissues to make sure the coins don’t
rattle around.
Place the three boxes in a pile on the table with the heavy one in
the middle of the pile (see Figure 10.1). Ask a volunteer to pick up the
boxes all at once to feel how heavy the pile is, then put them down.

75
76 Conjuring with Computation

Figure 10.1. First lift all three boxes, then just the top two.

Now explain you can change their weight at the click of your fingers.
Click your fingers and ask them to pick only the top two boxes up. They
will feel immediately heavier than the three did.
How can three boxes be lighter than only two of them?
Repeat this trick, having different people try it in turn.

The Magical Algorithm

TO DO Magically Weighted Boxes:
Before the trick starts
1. Fill one box with coins and tissues.
2. Place the boxes in a pile on the table with the heavy one in the
middle.
The trick itself
3. A volunteer picks up all three boxes together to see how heavy
they are.
4. The volunteer puts them back on the table.
5. Click your fingers saying you have changed their weight.
6. The volunteer picks up only the top two boxes.
7. Ask them to confirm these boxes feel heavier than all three did.

How It Works
It is all in the mind. It is like an optical illusion, but an illusion of touch
rather than of sight. Again, shortcuts made by our brains in making
sense of the world are being fooled. This is one example from a whole
 Magically Weighted Boxes: Your Brain Cannot be Trusted at All 77

family of size–weight illusions. Psychologists are still not totally sure

exactly why these particular illusions happen.

Thinking Computationally
Illusions can occur with all our senses, not just sight, though for similar
reasons. We really do have to accept that our brains have limitations
in the way they process information from the world. This matters when
designing multimodal interfaces that make use of other senses. For
example, increasingly computers are being built into physical objects so
that manipulating those actual objects in the world acts as an interface
to computers behind the scenes. Recognising the limitations of our brains
is not just about the way we design graphical user interfaces that are
based on sight.

Interlude: Lulu Hurst

Teenage magician, Lulu Hurst (1869–1950) was famous for tricks
that involved her showing off her apparently amazing superpower
strength. She claimed to have gained it superhero-style during a
lightning storm. In one trick, demonstrating how easily she could
overcome a massive weight, she had several men sit on a single
chair trying to hold it down. She would then show that despite the
weight and their efforts, she could easily move it. All her feats were
actually based on scientific principles. She retired at age 16.
This page intentionally left blank
 Chapter 11

The Teleporting Robot

Keep It Simple Stupid

Conjuring
You make a robot appear and disappear from a jigsaw as everyone
watches.

Computation
Interaction design principles are rules designers follow to help
ensure software is easy to use. One of the simplest rules is to keep
the interface simple. The more controls like buttons and drop-down
menus you have, the harder a system is to use. One way to keep it
simple is to avoid feature bloat. Rather than adding every feature
you can imagine, design just so that key tasks are supported and
designed to be easy to do.

The Trick
Download a copy of the magic jigsaw (see Figure 11.1) from
https://conjuringwithcomputation.wordpress.com

79
80 Conjuring with Computation

Figure 11.1. The Teleporting Robot magic jigsaw.

Cut out the six pieces of the jigsaw along the dashed lines. Complete
the jigsaw by placing the short piece with the grey robot in the top left
corner as in the original. With the audience, count the robots out loud,
and make sure everyone agrees there are 17. There is also a robot dog
and some Hexapus monsters (which are like an octopus but with only
six legs).
Now, mix up the pieces and then rebuild the jigsaw, but this time
with that first short jigsaw piece with the grey robot in the top right
corner. The other pieces still fit together but in different positions. You
are building it back with the pieces in the top two layers swapped over.
As you mix up the pieces and rebuild it, talk about how you have to keep
an eye on robots. As with WALL-E, they have a tendency to disappear on
their own little excursions once they start to develop a mind of their own.
Now, with the audience, count the robots again. To everyone’s
surprise, there are now only 16 robots. The 17th has gone, leaving no
trace. Ask which robot has disappeared and where it has gone. Did it
wander off? Did a Hexapus eat it, or did it teleport away?
Once everyone is sure that there are 16 robots, shuffle the pieces
and put them back the original way. Let everyone see there are 17 robots
again. Their challenge is to work out which robot disappeared and how
it did it.
 The Teleporting Robot: Keep It Simple Stupid 81

If they spot that you rebuilt the jigsaw in a different way, point out
that that explains nothing. It is the same pieces with the same heads
and bodies. They are just good recycling robots who swap their parts
rather than throwing them away. How can one disappear just because
they swapped bits?
If someone claims it is a particular robot that disappears, then have
them put a finger on the two parts of it, move the pieces around with
them pointing at the same two parts and they will see that both parts
are still there. That demonstrates their robot cannot be the one that
is gone. How could any part of a picture vanish from a physical jigsaw
anyway!

The Magical Algorithm

TO DO the Teleporting Robot:
1. Complete the jigsaw with the grey robot in the top left corner.
2. Count the robots out loud with the audience.
3. Note there are 17.
4. Mix up the pieces, then rebuild the jigsaw, this time with a green
monster in the top left corner.
5. Count the robots out loud with the audience.
6. Note there are 16.
7. Ask which robot disappeared and how.

How It Works
This works because of a limitation in the way our brains process what
we see. Look at the complete jigsaw. There are clearly 17 robots. When
you put the jigsaw together the other way, there are definitely only 16.
You’ve fooled yourself!
Error!!! Can’t Compute!
This trick is so powerful that you cannot see the robot disappear
even when you know what is happening. If you try and work out which
specific robot has gone, you have no chance of working it out. It’s not
one robot. All are involved.
82 Conjuring with Computation

If you have a hard problem to solve, then it is a good idea to simplify

it if you can. You may have noticed, for example, that the top four robots
are always there, unchanged. They just move position. That means they
are not involved. You can discard the top layer of pieces, and the trick
still works (just with 12 and 13 robots instead of 16 and 17). Try it. The
problem is now a little simpler.
How does it work? The original 13 robots are all that little bit shorter
than the 12 robots, something that is easy to miss. Bits of each of the
originals have been recycled into others to create a series of new, larger
robots. Each loses a small bit (a part of a head say) and gains a bigger
new version (a bigger head) from another robot. The first robot involved
is the one on the far left. It is completely on the bottom layer of the
jigsaw originally, but when the pieces are moved around, it has an extra
bit of head on top. That bit of head comes from another robot, which
loses that small bit of head but gets a bigger bit in its place. As you
move from robot to robot, the place they are cut moves down their
body. Eventually, in the final robot in the sequence (the one above the
dog), the whole robot starts on the middle piece, and when it is moved,
it leaves a gap behind. It also gets taller by joining on to some legs on
a bottom piece. All 13 robots along the bottom have squashed together
into 12 new taller robots.
We can simplify this even more to help see what is happening by
replacing the robots with lines and arranging them so they swap with
the adjacent line. Look at the picture of the lines in Figure 11.2.
There are five vertical lines on the left. Now imagine you cut those
lines along the black diagonal line and slide the top triangle along by one
space. The result is four lines as on the right (draw some parallel lines
and try it). Where did the extra line go? It can’t just have vanished,
can it? No, what’s happened is much easier to see here: the final lines
are all a little bit longer. The missing line has just combined with part
of another line. That other line’s remainder combined with part of yet
another, and so on.
We’ve made it more obvious here by only using a few lines and
making the differences larger. Why don’t we spot it with the robots?
Well, people aren’t that good at accurately judging small length
 The Teleporting Robot: Keep It Simple Stupid 83

Figure 11.2. A simplified version of the effect used in the magic jigsaw. Slide the
top half so each line joins with the bottom of the next along so becomes taller. You
are left with one fewer line.

differences by eye. The new robots look “about the same” as before.
“About”, yes, but “the same”, no! That’s why we invented rulers!
The robot jigsaw design is more cunning than the lines though.
Rather than just having the robots in order, where a simple slide of the
top section would make the remaining robots taller, the jigsaw design
has lots more confusing detail for our brains to take in but also weaves
the vanish across different parts of the picture. That makes it hard to
take in everything that you need to see in order to see what is actually
happening even when you know.

Presentation matters
Magic is about more than algorithms, about more than the self-working
part. You have to get the presentation right too.

Magic = Algorithm + Presentation

Even the most cunning trick can fail to be magical without a good
presentation. The two versions of the trick, one with the lines and one
84 Conjuring with Computation

with the robots, use the same mechanism. It is the presentation that
makes one impossible to see what is happening, whereas in the other it
is fairly easy. Here it is done by making the picture more complicated.
There is too much going on for anyone’s brain to take it all in.
Magicians rely a lot on understanding how our brains work, and
particularly how they get things wrong. They rely on an understanding
of cognitive psychology: the science of how we think. The key point is
that our brain has limited resources. For example, we can only pay close
attention to one place at once and only remember a few things at once.

Thinking Computationally
Presentation matters
Computer science, and especially writing software, is about more than
algorithms too. You have to get the presentation right, just like a
magician.

Software = Algorithm + Presentation

The difference is that a magician makes sure the whole audience

makes the same mistake at the same time, whereas when using programs
no one should make a mistake as we want programs to be easy to use.
Computer scientists need to understand the same cognitive science as
magicians, just use it in reverse.
For example, think about the remote control for a TV. It is very
complicated with lots and lots of buttons. If someone accidentally presses
the mute button switching the sound off, it can be quite hard to find
the right button to get it back on again, unless you use it a lot. There
is too much information there for our brain to take in.
One thing designers do to help make sure they create usable
interfaces is to follow design principles. Our teleporting robot trick and
the problems people have with remote controls lead to the “KISS” design
principle: Keep It Simple Stupid! The best interfaces, that are easiest
to use, are often very simple. They aim to support one task at a time
and do it well. The things you might want to do at any given time are
always really obvious. One thing that helps is to avoid feature bloat.
 The Teleporting Robot: Keep It Simple Stupid 85

Rather than have your software try to do everything for everyone, focus
on designing in a way that makes it easy to do the key tasks one group of
people want to use it for. If other people want to do completely different
things, then you design a different interface for them.
The reason many people think they are not good at using computers
and other gadgets is not because those people are stupid. It is because
the people who designed the software or gadget did not think hard
enough about people. They made the programs work more like a magic
trick than a usable program, so hard to follow and easy to get wrong. If
you become a programmer or interface designer, make sure your designs
are really easy to use, so no one makes mistakes. Make sure they are the
opposite of a magic trick.

Interlude: Sam Lloyd and

William Hooper
The teleporting robot trick is based on a very old mathematical
puzzle. A very simple version of it appears in The Moscow Puzzles
(1972) by Boris A. Kordemsky where a line is made to disap-
pear. A fairly early and famous version of it was by Sam Lloyd
(1841–1911), the great inventor of many mathematical puzzles,
though he didn’t invent this one. An even earlier version is described
by William Hooper in a book published as far back as 1794. It
involved cutting up bank notes. It gave a way to cut nine notes
up into 18 parts. They were then put back together again to make
10 banknotes. Modern bank notes are now always designed in a way
that stops the trick from working!
This page intentionally left blank
 Part 4

Decomposition and Abstraction

When inventing new tricks, we can build them out of a collection of

basic parts. Such parts can be reused across other tricks. Likewise, we
can solve the problem of writing a program by breaking it into parts to
write separately or just reusing bits previously written.
This page intentionally left blank
 Chapter 12

Pre-booked Picture Magic

Abstraction

Conjuring
By reading the mental impression of it in their head, you are able
to draw a picture that is selected randomly from a book of images
by an audience member.

Computation
An abstraction is a simplification of something. Computer scien-
tists use abstraction a lot, both as a way of making a complex piece
of software easier to understand and to help make it possible to
write in the first place. Similar ideas can be used in magic.

The Trick
You pass a small notebook around the audience for them to inspect.
In it, you have written, on numbered pages, a series of random objects
that came to you “in a dream”. The spectators check that each page
does have a different and unrelated thing written on it. The task is now
to choose one.

89
90 Conjuring with Computation

You use a pack of cards to ensure a random choice. Spread the

cards at the start to show the random mix. A volunteer then cuts the
cards. They add the values of the bottom two cards of the pack after
the cut. Remind them that a Jack is worth 11, a Queen is worth 12 and
a King, 13. They secretly note this number. The spectator then takes
the notebook to the other side of the room and goes to the page number
given by their random card-created secret number. They then close their
eyes and visualise the object written on that page.
On the other side of the room, you, the magician, sketch on a pad
your mental impression of the thing the spectator is thinking about. The
spectator then reveals the identity of the thing on that randomly chosen
page. You turn your pad around showing you were able to correctly read
the spectators’ thoughts from a distance and draw that object.

How It Works
This effect relies on two separate elements: a mathematical technique
called the 14/15 stack, used to force page numbers, and the notebook
itself.
To create the stack (before doing the trick), you must first
remove the Ace of Hearts and the Ace of Spades from a normal pack of
cards. The fact that this pack only has 50 cards rather than 52 won’t
be spotted. You now need to carefully arrange the remaining cards in
the following full-deck stack order (where C means Clubs, D means
Diamonds, H means Hearts and S means Spades). It should end up with
the 7 of Diamonds at the bottom of the deck, with all the cards face
down.
7C 8C 6D 9S 5C 10H 4D JS 3C QS 2D KS AC KH 2H QC 3D JH 4S
10C 5D 9C 6S 8S 7S 8D 6H 9H 5H 10D 4C JD 3S QH 2S KD AD KC
2C QD 3H JC 4H 10S 5S 9D 6C 8H 7H 7D
This order is specially created to ensure that every adjacent pair
of cards, including the last and the first, add up to either 14 or 15.
Cutting the pack does not actually change the cyclic order, it just starts
it in a different place. That means when the volunteer cuts the pack,
the bottom two cards of the resulting pack will still add up to either
page 14 or 15.
 Pre-booked Picture Magic: Abstraction 91

Figure 12.1. The picture you draw.

You now have a way to use a free cut of the pack to force either
page 14 or 15. You next need a way of dealing with the problem that
there are still two possible outcomes. That’s where the way you set up
the notebook and make the prediction comes in.
On your prediction pad, you sketch a large circle, with a smaller
circle in the middle (see Figure 12.1). This prediction and its simple
design allow you to claim it represents either of the two things you write
on pages 14 and 15. Use your imagination for the objects you name on
pages 14 and 15 of the book. For example, you could write Doughnut,
Car Tyre or Porthole. That means you are covered whichever of the two
values the spectator uses.
Remember that the spectator will tell you their choice before you
show your prediction, so you can be ready with the right convincing
argument about how the mental impressions you got made you roughly
but correctly draw the answer they gave you. If you really could read
minds, it would be difficult, so the fact you didn’t get the sugar sprinkles
on the particular doughnut they were visualising will be forgiven.
Use a cheap spiral-bound notebook for the book, as it allows for
easy flipping of the pages. you need to create the illusion of free choice.
Therefore, fill the book with enough random and unrelated things to
make it convincing. It should certainly include a lot more than 15 things.
Write the names clearly in block capitals so they are easy to read and
92 Conjuring with Computation

clearly label the pages in the same place at the bottom, as they need
to easily find the page number. Importantly, make sure that you use
a separate sheet of paper for each object. Don’t write the things on
facing pages in case your spectator gets suspicious when they see that
the outcomes on both pages 14 and 15 could match your prediction.
There are lots of alternative presentations for this trick. You could,
for example, use world landmarks, either writing the names of the
landmark or creating a book with pictures of the famous sites. Then
your prediction could be an inverted V shape pointing upwards and a
line underneath signifying the ground. The landmarks on pages 14 and 15
could then be The Eiffel Tower and Mount Everest. Both are arguably
represented by your prediction. Come up with your own version: get
creative!

The Magical Algorithm

TO DO Pre-booked Picture Magic:
Before the trick starts
1. Take a spiral-bound notebook and number every second page in
the bottom corner, starting from 1 and going up to at least 35.
2. On pages 14 and 15, write the words DOUGHNUT and CAR
TYRE, respectively, in block capitals.
3. On each other numbered page, write a noun in large block
capitals.
4. Take a pack of cards and discard the Ace of Hearts and the Ace
of Spades.
5. Place the remaining cards in the order: 7C 8C 6D 9S 5C 10H
4D JS 3C QS 2D KS AC KH 2H QC 3D JH 4S 10C 5D 9C 6S
8S 7S 8D 6H 9H 5H 10D 4C JD 3S QH 2S KD AD KC 2C QD
3H JC 4H 10S 5S 9D 6C 8H 7H 7D where C, D, S and H mean
Clubs, Diamonds, Spades and Hearts, respectively, and A, J, Q
and K mean Ace, Jack, Queen and King, respectively.
The trick itself
6. Ask a volunteer to cut the pack, completing the cut by placing
the bottom pile back on the top.
 Pre-booked Picture Magic: Abstraction 93

7. Have the volunteer take the bottom two cards and add them up,
remembering the total.
8. Give them the book.
9. The volunteer goes to the other side of the room, finds the page
of their remembered number and silently reads the word on that
page.
10. The volunteer closes their eyes and visualises the object given on
the chosen page. While they do so, you pretend to focus on their
thoughts, telling them you will draw what is coming through to
you from their thoughts.
11. Draw two concentric circles on your pad.
12. The volunteer tells you the word they were thinking of.
13. Show them what you have drawn and point out you drew the
object they were thinking of.

Thinking Computationally
Sketches are an abstraction, a simplification of the real thing. Your
prediction is an even simpler version still. Abstraction is all about giving
a simplified version of things, but that still captures the essence of
the original. The abstraction (like the drawing) discards detail. A good
abstraction will only throw away detail that is unnecessary for the task
at hand.
Computer scientists use abstraction a lot, though they have very
precise versions of it, that we look at in more detail in later chapters.
For now, think about it as hiding or discarding detail that is unnecessary
for the task at hand (a doughnut is still recognisable as a doughnut even
without the sugar sprinkles).
The most important kinds of abstraction we will look at are about
program instructions themselves, but computer scientists do use this kind
of drawn abstraction in one situation: icons. Icons are small pictures used
to represent programs. The image is a simplified picture representing
what the program is or does. For example, a trash can represents the
deletion operation on most graphical user interfaces. Drag a file to the
trashcan picture and it is deleted. Icons are very simplified versions of
the thing they represent, but to work well they must catch the essential
94 Conjuring with Computation

detail and be instantly recognisable. The best are really good visual
abstractions.
The magic trick, by contrast, relies on a poor abstraction in the
sense that a pair of circles can represent two different specific nouns in
the book (though it is a perfect abstraction for its purpose: to make
the trick work). Graphical user interfaces need much clearer ones than
this. If you could mistake an icon for meaning two different things, it
would be a poor icon. Many modern icons are very confusing in this way.
Abstractions generally lose information, but a good abstraction will not
lose anything important!

PAUSE: Susan Kare

Susan Kare was part of the team working at Apple on their first
graphical user interface. She created the first icons, and her work
and ideas were key to the success of Apple both then and since.
With no direct experience of Pixel Art to draw on, she instead
gained inspiration from pixel-like art of the past: Seurat’s pointillism,
needlepoint and mosaics. Despite the icons being very small, they
were simple and clear, following her design maxims of “meaning,
memorability, and clarity”. She was responsible for Apple’s visual
design language designing fonts as well as icons.
 Chapter 13

Trained Rice
Decomposition and
Procedural Abstraction

Conjuring
You train rice so that unlike normal rice it can grip onto a
chopstick.

Computation
Programs that are millions of lines long cannot be written without
ways to simplify the task. The most immediate way is to use
decomposition: split the big difficult program into lots of small
meaningful parts that can be treated as distinct problems (i.e., mini
programs) to solve. These separate parts are called procedures. By
giving each a name, we can use them later just by giving their
name.

The Trick
For this trick, you need two small plastic drink bottles with narrow necks,
chopsticks and uncooked long-grain rice.

95
96 Conjuring with Computation

Figure 13.1. The bottles are filled with rice and a chopstick is inserted.

You must prepare the bottles by filling them with long-grain rice
before you present the trick. Fill the first with rice and leave it as is.
Pushing a chopstick into it, the chopstick should go in and then come
out easily (Figure 13.1).
With the second bottle, after pouring in the rice, repeatedly tap the
bottle on the table so that the rice packs closely together. As you do,
you will create more space at the top. Keep adding more rice to fill it up.
When it no longer packs down, try pushing a chopstick into it. It
should be harder to do than with the first bottle. If it is ready then,
when you pick the chopstick up, the whole bottle should lift and stay in
place. If not, keep tapping the bottle packing the rice even closer.
Remove the chopsticks, and you are ready.
Now for the performance. Announce that you have discovered a
strain of rice that can be trained to grip things. Show the two bottles
and explain that you have normal rice in the first bottle but trained rice
in the second.
 Trained Rice: Decomposition and Procedural Abstraction 97

Place the chopstick in the first bottle and demonstrate that it does
not grip the stick. Allow a member of the audience to try. Now place a
chopstick in the second bottle and moving your mouth up close, whisper
to it over and over “grip the chopstick, grip the chopstick. . .” Then,
slowly lift the chopstick and watch the rice grip and the bottle rise
with it.

The Magical Algorithm

Here is the trick written out as a series of steps in a structured way similar
to that of the previous tricks. A difference to our previous instructions,
though, is that here we have broken (or decomposed) the whole trick
into two additional smaller named steps.
TO DO Trained Rice:
1. DO the preparation of trained rice.
2. DO the performance of trained rice.
TO DO the Preparation of Trained Rice:
{This should be done before the show.}
1. Fill two small plastic drink bottles with rice.
2. Take one of the bottles.
3. DO
a. Tap the bottle gently on the table to make the rice settle.
b. If more space appears, refill it with rice.
c. Push a chopstick into the rice.
d. Try to pick the bottle up with the chopstick.
e. Remove the chopstick.
REPEAT THE ABOVE UNTIL the chopstick stayed in place when
picked up.
TO DO the Performance of Trained Rice:
1. Announce that you have discovered a strain of intelligent rice that
you have trained.
2. Bring out the first (untapped) bottle, explaining it is normal rice.
3. Place a chopstick in the rice.
4. Try to lift up the bottle just with the chopstick. You can’t.
5. Bring out the second (tapped) bottle, explaining it is trained rice.
98 Conjuring with Computation

6. Whisper to the second bottle.

7. Push the chopstick into the second bottle.
8. Lift chopstick and bottle.

How It Works
The secret of this trick is just the laws of physics and, in particular,
friction. In the first bottle, the rice is not tightly packed so the grains
easily slip past each other. In the second because they are pushed tightly
against each other and the edge of the bottle, there is more friction
between the rice grains and between the rice grains and the chopstick.
This friction is enough to hold the chopstick in place.

Thinking Computationally
Here we have decomposed the algorithm into two subparts: the
preparation and the performance. In fact, we did that informally in earlier
tricks, though without naming the parts and without giving separate
explicit instructions of the order to do them. Decomposition is just the
idea of breaking up an algorithm into smaller, easier-to-manage parts
like this.
Here, we are using decomposition and a programmer’s version of
abstraction (referring to the parts by name) purely as a way to organise
the way we write the algorithm. This is to make it easier for a human
(a magician or a programmer) to understand the instructions.
This matters a lot when writing large programs. Writing programs is
about writing instructions computers can follow. Writing good programs
is also about making them easy for humans to understand, so follow too!
If you don’t, then they are likely to have more bugs and, when changes
to the program are needed, those changes will be harder to do.
Decomposition into meaningful parts is one of the most important
ways of making programs easier for humans to understand. Often
programs or even individual algorithms have a setup part, where data is
collected or organised, followed by the part that does the actual work,
and this is usually a good way to decompose the program. Both parts
have very clear tasks to do. What you don’t want to do is decompose
 Trained Rice: Decomposition and Procedural Abstraction 99

a program you are writing into arbitrary parts where the separate parts
have little real meaning.
There are three important aspects to what we are doing here.
The first is decomposing the algorithm into parts. The second is to
give those parts a unique name that can be used to indicate when
those instructions should be followed. The final aspect is to use that
name instead of the actual detailed instructions when we do wish it to
be followed. These latter points combine to give another key tool in the
computational thinking toolkit, that of abstraction or hiding of detail.
We previously saw abstraction in drawings. Here, we are doing a more
technical kind of abstraction. This naming of parts of an algorithm is
called procedural abstraction. It involves using a name in place of a
group of detailed instructions (hiding the detail of the instructions behind
the name). These groups of instructions are known as procedures. We
will explore procedural abstraction more in later chapters.

A Magical Diversion
Magicians can think of the performance of great tricks as being
decomposed into three parts1 in addition to the preparation:

The Pledge is where you set the scene, giving the effect its meaning;
The Turn is where you perform the trick leading them along with
the story;
The Prestige is where you reveal the twist at the end, the delight!

For example, we could decompose the Trained Rice performance in

this way:

TO DO the Trained Rice Performance:

1. DO the Trained Rice Pledge.
2. DO the Trained Rice Turn.
3. DO the Trained Rice Prestige.

1
Named by author Christopher Priest in his novel, The Prestige (1995).
100 Conjuring with Computation

TO DO the Trained Rice Pledge:

Announce that you have discovered a strain of intelligent rice that
you have trained.
TO DO the Trained Rice Turn:
1. Bring out the first (untapped) bottle, explaining it is normal rice.
2. Place a chopstick in the rice.
3. Try to lift up the bottle just with the chopstick. You can’t.
4. Bring out the second (tapped) bottle, explaining it is trained rice.
5. Whisper to the second bottle.
6. Push the chopstick into the second bottle.
TO DO the Trained Rice Prestige:
Lift chopstick and bottle.
By working with this standard structure, this standard decomposi-
tion, you improve the chances your trick will seem magical. For example,
missing out the pledge in the above would mean it works less well as
a trick even though the mechanics still work. If you are inventing your
own tricks, make sure they have pledge, turn and prestige!

PAUSE: Michael Faraday

Michael Faraday (1791–1867) was a self-taught scientist and inven-
tor who also created the inspirational Royal Institution Christmas
Science Lectures that still run for children today. He delivered 19
of them personally. He showed that science could be taught to the
general public in an entertaining way. He was a critic of spiritualism
and seances and ran scientific experiments on how certain seance
effects were achieved. As a scientist, his achievements were many
across multiple fields including chemistry, physics and environmental
science, but he is best known for his work on electromagnetism, with
inventions such as the dynamo that ultimately made the widespread
use of electricity a reality.
 Chapter 14

The Teleporting Top Hat

Hiding Detail and Swapping Parts

Conjuring
We develop a new version of the Invisible Palming trick, replacing
the palming part with a top hat teleporter.

Computation
Decomposition, combined with procedural abstraction, allows us
to build large programs from small. Once a procedure is written,
we can forget about the detail of how it is actually implemented.
This also means that we can later completely change the way a
step is done without it affecting the rest of the code that uses it.

The Trick
This trick is exactly the same as Invisible Palming (Chapter 1) except
that you swap in new steps for those where you pretend to move the
card from one pile to the other. Instead, you “move” the card with
your magical top hat, into which the card disappears and reappears.
At the point where you are to move the card, place your top hat over
the pile with the extra card. Tap it with your magic wand and say the

101
102 Conjuring with Computation

magic words. Move the hat to the other pile, tap it once more with your
magic wand and say the magic words once again.

The Magical Algorithm

One step of the earlier Invisible Palming trick was to pretend to move a
card from one pile to another (no actual movement was needed). The
algorithm for it might be split off as follows:
TO DO Pretend to Move a Card Between Piles:
1. Place a hand over the pile with the extra card.
2. Rub the back of that hand.
3. Lift your hand, showing the palm before placing it over the other
pile.
4. Tap your hand and remove it, saying the card has moved.
This is then used as one step (step 4) of the whole thing as follows:
TO DO Invisible Palming:
1. Deal out 15 cards in pairs with one left over.
2. Take the pairs back splitting them into two piles.
3. Have a volunteer choose where to put the extra card.
4. DO Pretend to Move a Card Between Piles.
5. Reveal that the extra card has magically moved piles.
Now we can invent our own presentation of the trick (no longer invisible
palming), just by changing the instructions of this one procedure. The
above instructions do not need to change though.
TO DO Pretend to Move a Card Between Piles:
1. Place your top hat over the first pile.
2. Tap it with your magic wand, saying “Abracadabra”.
3. Move the top hat to cover the other pile.
4. Tap it with your magic wand saying “Abracadabra”.
5. Say that the card has moved.
The rest of the trick does not change, so the instructions do not need
to be changed. This is one way that new tricks are developed: each
 The Teleporting Top Hat: Hiding Detail and Swapping Parts 103

magician creating a personal version. Don’t copy, do your own version,

and do better!

How It Works
The mechanics of the trick are exactly the same as they were before.
We have actually just changed the presentation of one part. It doesn’t
change how or why the trick worked. The substituted step does exactly
what the original did (nothing to the cards, but a lot to make the
audience think something else is going on).

Thinking Computationally
We have used our procedural abstraction (giving a group of instruc-
tions a name), to make it easy to swap a new implementation for one
of the steps, i.e., one of the procedures. The key point is that the main
instructions do not change at all. As long as the new steps have the
same effect as the old, nothing else has changed.
The same can be done with programs. If the program is split into
procedures, then we can take any of those procedures and change how it
works as long as we do not change what it does. The rest of the program
is then guaranteed to still work, without any change to anything but the
instructions in that one procedure.
The use of procedures makes a program much more maintainable
because of this. It is easier to improve. It is simple to change the
implementation of this part of the program in that when doing so we
do not have to worry about any of the rest of the program. We know
nothing else has to be changed. We might want to do this, for example,
to make a critical part work faster or use less storage space.
Furthermore, if the procedures are used multiple times, improving
the procedure improves it everywhere it is used throughout the program
with no extra effort: we do not need to track down all the places it is
used as no changes there are needed.
104 Conjuring with Computation

PAUSE: Jeannette Wing

A computer scientist who has also become a martial arts expert,
Jeannette Wing has worked mainly in the area of using maths-
based specification languages to very precisely state what programs
should do. However, her biggest contribution has been to promote
the idea of computational thinking as a core set of general skills
that computer scientists should, and do, learn. The term was first
introduced by Seymour Papert many years earlier, but she gave it life.
She argued that the skills and way of thinking were generally useful
in life, not just for writing programs, and so should be taught to
all. As a result, when computer science and programming started to
be introduced in school syllabuses around the world, computational
thinking was at their core. Her students gave her the nickname
“Dragon Lady” because she was a tough teacher. The name then
stuck once she started doing martial arts.
 Part 5

Procedures and Procedural Abstraction

There are many different kinds of false shuffles and similar well-defined
magical manoeuvres that have a specific effect. Just by giving the name
of one of these steps, a magician would know what was meant and be
able to use it as part of a trick.
This corresponds to procedural abstraction where clear self-
contained pieces of code are created for use in lots of programs. Once
created, they are incorporated into any running program (or “called”)
with a command that just gives their name.
We describe a series of shuffles and cuts. In some, you really do
mix the pack; in others, you just appear to shuffle it, but actually leave
either the whole pack or an important part of it, unchanged.
This page intentionally left blank
 Chapter 15

Overhand Shuffle
Procedures

Conjuring
This shuffle really does mix up the cards in a random way.

Computation
When we have a meaningful task that we will want our instructions
to refer to lots, we package the instructions for that task as a
procedure, give it a name and refer to the name, using it as an
instruction (call the procedure) at the point we want the task to
be done.

How to Do the Shuffle

The overhand shuffle is a commonly used, good way to mix cards. The
deck is cupped in one hand (right if you are right-handed, left if left-
handed) between the thumb and fingers, and the cards are transferred
to the other hand by peeling off a small packet of cards from the original
deck with the other hand’s thumb. These small piles are deposited at
the front of a new deck forming in the second hand. This is repeated
until all the cards in the first hand are mixed into a pile cupped in the

107
108 Conjuring with Computation

second hand. The whole thing is then repeated several times until the
cards are thoroughly shuffled.
It is a legitimate shuffle and does mix the cards. However, there
are various ways it can be performed that keep parts of the pack in the
places you want them while still looking like a real shuffle, as we will see.
If you haven’t done an overhand shuffle before, practice it for a while,
and once you are reasonably good at it, you can start to get sneaky.

The Magical Algorithm

This algorithm can be done repeatedly for more thorough shuffling.
TO DO an Overhand Shuffle:
1. Cup the pack in one hand between thumb and fingers.
2. DO THE FOLLOWING UNTIL all cards are in the other hand.
a. Pull off a small packet of cards from the front of those in
the first hand with the other thumb into the front of the pile
forming in the second hand.
3. Transfer the cards back to the first hand.

Thinking Computationally
The overhand shuffle is a generally useful card algorithm. You might
use it in a range of tricks where the order of the cards doesn’t matter
at the start. We don’t want to have to explain it in the instructions to
tricks every time we need it though. Instead, in the English description,
we might just say “Start by shuffling the cards”. More precisely and
rigorously, the first step of the trick’s algorithm might just be:

1. DO an Overhand shuffle.

We rely then on the instructions about how it is done to be

somewhere else that is accessible (possibly in the magician’s head).
When we originally give the instructions, we define the procedure: we
are writing a procedure definition. In this book, the overhand shuffle is
defined as this part’s “Magical Algorithm”. That allows us to refer to it,
 Overhand Shuffle: Procedures 109

and you, the reader, now know precisely what we mean, anytime we
mention its use!
The definition itself is not telling you to do the trick now (though
you may of course want to practice it and check it works now). It is not
a trick on its own. Elsewhere you will be told to do it when needed with
an instruction like the DO instruction above as part of an actual trick.
That is the point when the instructions are followed. Programmers refer
to it as the procedure call.
This is decomposition and abstraction being used in tandem again.
We have split off a part of the instructions and given them their
own name. The difference here to our previous use is that we have
decomposed our trick in a way that gives us a generally useful procedure
that will be used over and over again in lots of tricks, and possibly even
several times in the same trick. The benefit is thus much more than
organising that one trick. At a stroke, we have set up a decomposition
to be used in tricks that we haven’t even thought of yet. Whenever we
are planning a new trick and need a shuffle, we no longer need to worry
about how it is done at all. We do not need to think about the algorithm
of the shuffle, just know what it does: i.e., mix up the cards.
This is a really important use of decomposition and procedural
abstraction. When deciding how to decompose, one important question
to ask is: “can I create a generally useful procedure here.” If you can,
it is likely to be a good decomposition. In the software world, classic
“library” procedures like this are often provided ready written to include
in your programs as predefined procedures. Examples include code to
search a list of data, sort data into numerical order, print a message to
the screen, read some data from the keyboard, write to a file, and so on.
The idea is used everywhere in programs. Most languages include lots
of libraries of previously written, predefined procedures. Like our shuffle,
the instructions can be written down once, made sure they always work,
and then referred to by name, in later programs we, or other people,
write. When writing instructions that use them, we no longer have to
think about how that code actually works, we just need to know what
it does so that it is right for our current purposes.
110 Conjuring with Computation

Interlude: Persi Diaconis

Persi Diaconis was a professional magician with a passion for
debunking crooked casino games. This drew him back to the
advanced maths he vowed as a school dropout that he would one
day master. He is now a professor of Mathematics and Statistics
at Stanford University where he studies the randomness in events
such as coin flipping and shuffling playing cards. He and a fellow
mathematician, David Bayer, proved that to randomise a pack of
cards, you need only seven riffle shuffles where you cut the pack and
then combine them by interleaving cards.
 Chapter 16

Swapped Ends Overhand Shuffle

Specifying Procedures

Conjuring
This shuffle leaves the pair of cards that are at the top and bottom
of the pack in place, just swapped.

Computation
It’s important that it’s clear what each procedure does, so we
know when to use it. This can be done with English comments
or using maths. Important things to be clear about are both what
it ultimately does and any assumptions that must hold before the
procedure is called if it is to work correctly.

The Trick
Take the deck in your right hand for a standard overhand shuffle. Peel
the top card off as a single card, depositing it in the empty hand
(Figure 16.1(a)), then proceed normally pulling piles of cards off the
pack in a standard overhand shuffle way and dropping them in front
of that first peeled card. Continue doing this until you get near the

111
112 Conjuring with Computation

(a)
(b)

Figure 16.1. To do a Swapped Ends Overhand Shuffle, (a) first peel off the single
top card and then drop the rest of the pack as in a normal shuffle in front of it. (b)
After doing a normal shuffle for the middle of the pack, leave the last card on its own
to be dropped to the front in the last step.

end of the pack. At that stage, pull the last few cards off one by one.
In particular, peel off the very last card on its own (Figure 16.1(b)).

The Magical Algorithm

We add a specification to this algorithm. In curly brackets, we give
any condition that must hold for the algorithm to work at the start.
We also state what will have become true after the algorithm has been
completed at the end.
TO DO a Swapped Ends Overhand Shuffle:
{Nothing is assumed.}
1. Cup the deck of cards in one hand between thumb and fingers.
2. Peel off the top card into your other hand.
3. DO THE FOLLOWING UNTIL you are near but not at the end
of the deck.
a. Pull off a small packet of cards from the front of those in
the first hand with the other thumb into the front of the pile
forming in the second hand.
4. DO THE FOLLOWING UNTIL all the cards are in your second
hand.
 Swapped Ends Overhand Shuffle: Specifying Procedures 113

a. Peel off a single card with the second hand’s thumb into the
front of the pile forming in that hand.
{The first and last cards have swapped positions.}

How It Works
The first step moved the top card to the bottom. Peeling the last card
from the stack alone means that the final card, which was the original
bottom card, is now deposited on the top. Cards in the middle have
been mixed a little. The top and bottom pair of cards are still at the top
and bottom, just swapped over.

Thinking Computationally
An important property of well-written procedures is that they have a
clear and clean specification. That is you can easily and precisely say
what they do (as opposed to how they do it). For example, the algorithm
of this shuffle guarantees, whatever the initial order of the pack, to swap
the positions of the first and last cards. That is the specification of the
algorithm. Nothing is guaranteed about any other cards. That is left
unspecified.
Specifications come in two parts. A precondition (what must be
true before the procedure is followed for it to work) and a postcondition
(what will then be true afterwards). The precondition here is basically
that the cards can be in any order (though it is the case that cards of
interest are in the first and last positions). The postcondition is that
those two cards have been swapped.
A precondition says what must be true for it to work. We have
specified that anything can be true by writing “Nothing is assumed”.
Computer scientists specify this by just writing “true” (stating that all
that matters is “true must be true” which is always the case so not
imposing any restrictions!).
Details matter though, and actually here perhaps there could be a
precondition. We have referred to the deck of cards in the algorithm, and
an implicit assumption is that this is a normal deck. The shuffle works
114 Conjuring with Computation

with any cards so if we wanted to take that into account explicitly, then
something to note as a precondition is there have to be at least two
cards in the deck. The precondition could then be

{The deck of cards contains at least two cards.}

Doing it with two cards wouldn’t be very magical though, so actually

the number in the precondition ought to be larger for a magical effect.
When writing programs, it is equally important to be clear what the
preconditions and postconditions of a procedure are. What the procedure
does should also be easily described in English. If it isn’t, then it probably
wasn’t a very well-thought-out procedure. For example, a procedure
might find the position of a given number in a list. The precondition
might be that the list of numbers is in ascending order. However, English
is not really precise enough to write specifications, just as it isn’t precise
enough to use to write programs.
Well-designed programming languages, used for writing critical code,
give a way of writing specifications formally in the program as what are
called assertions. Rather than being written just in English, as with
other aspects of the language, they are written in a precise language
with precise meaning.
For the purposes of the magic trick, we could also give a more
complex postcondition saying what the aim is in terms of what the
audience believe:

{The first and last cards have swapped positions AND

The audience believe the deck is in a random order.}

Of course, we can’t be certain of what the audience believe, so such

claims are not mathematically checkable results in the same way as the
positions of cards. However, if our aim was to make clear the aim of
the trick, this would make sense. In fact, specifications about beliefs
are used to reason about algorithms called security protocols, where
the aim is to check that a hacker cannot gain access to a secret (like a
password) when the algorithm is followed.
 Swapped Ends Overhand Shuffle: Specifying Procedures 115

PAUSE: Nikola Tesla

Nikola Tesla (1856–1943) was a prolific inventor of gadgets using
electricity and electromagnetism, but also a showman. He held public
demonstrations of his ideas and inventions. One such demonstration
in the 1890s was of wireless-controlled lighting, where he lit up
lights on the other side of a stage. It could just as easily have
been presented as a magic show! He predicted similar systems could
be used for wireless communications and created the world’s first
wireless-controlled boat.
He is most famous, though, for his inventions making alternating
current (AC as opposed to direct current, DC) a practical reality. It
is now used for long-distance power transmission. His AC induction
motor used a rotating magnetic field. He created the “Egg of
Columbus” demonstration to illustrate this. It used a magnetic field
to make a giant egg made of copper stand and spin on its end.
Tesla appeared as a character in Christopher Nolan’s film about
rival magicians, The Prestige. He was played by David Bowie, chosen
because the part needed someone who was “extraordinarily charis-
matic”. Rock star Bowie incidentally also invented and co-wrote one
of the first programs to help people be creative, helping him write
lyrics.
This page intentionally left blank
 Chapter 17

Ends Overhand Shuffle

New Procedures from Old
and Layers of Abstraction

Conjuring
This shuffle leaves the pair of cards that are at the top and bottom
of the pack in their original place.

Computation
Once you have created generally useful procedures and made sure
that they work, you can start to build new procedures out of them.
You build new reliable code from existing reliable code, rather than
creating everything from scratch.

The Trick
In this trick, you just do the Swapped Ends Overhand Shuffle twice.

117
118 Conjuring with Computation

The Magical Algorithm

TO DO an Ends Overhand Shuffle:
{You have a normal deck of cards.}
1. Do the Swapped Ends Overhand Shuffle.
2. Do the Swapped Ends Overhand Shuffle.
{The deck is shuffled, but the first and last cards are unchanged.}

How It Works
As a single Swapped Ends Overhand Shuffle swaps the places of top
and bottom card, doing it twice puts them back to where they started.
Cards in between change position.

Thinking Computationally
Once we have created a procedure, we can use it as though it was a
core, primitive instruction provided by the language. In particular, we
can build new, generally useful procedures out of simpler ones as here.
From the original false shuffle, we create a new false shuffle that does
something slightly different (leave the first and last cards alone). We are
building layers of abstraction. First, we create primitive procedures.
We then no longer need to worry about how they work when designing
more complex procedures. As long as we know our basic false shuffle
swaps the first and last cards, we don’t care how. That knowledge is
enough to be sure that doing it twice will leave the first and last cards
alone. Now, knowing exactly what our new shuffle does, we can use it
when designing new tricks, without worrying how the original basic step
or the new composite one is actually done while inventing the new trick.
The details of lower layers of procedure are hidden in the instructions of
later ones (see Figure 17.1). This layering of procedural abstraction is
key to creating really large programs that do actually work. Without it,
the detail quickly becomes too complex to avoid mistakes.
Really well-designed programs are built out of a set of primitive
procedures. A more complex set of primitives are created from those,
Ends Overhand Shuffle: New Procedures from Old and Layers of Abstraction 119

Figure 17.1. Building new components from existing ones in layers of abstraction.

then more complex ones from those, and so on. Groups of such
primitives are put into libraries to be incorporated into any program
that needs them.
Of course, we are talking here about the process of designing new
tricks (and programs). You do, of course, need the detail when actually
doing a trick using such a shuffle, and those details are there, in the
named instructions, when you need them. When practising a trick from
a book like this that needs a false shuffle, for example, you just jump to
the right page for each step, and when done jump back.
However, this multilayered decomposition probably helps a lot to
learn a trick too. You can first learn the basic false shuffle. Only when
you have got that down perfectly, and the trick is “loaded” into your
brain, do you start to practise the more complex version that uses it
twice. Thus this sort of abstraction, this computational thinking trick, is
helpful for humans following instructions, not just when writing programs
or designing instructions.
120 Conjuring with Computation

Interlude: Howard Thurston

Howard Thurston (1869–1936) was one of the greatest magicians
of the late 19th and early 20th centuries. He marketed himself as
the “King of Cards” based on his skill at card tricks. This included
making cards appear and disappear from his fingertips and a trick
where cards chosen by members of the audience rise in turn to the
top of a pack placed inside a glass. As a boy, he was more street
urchin and con artist than his later sophisticated persona might
suggest, travelling from city to city by hiding on freight trains. He
tried to become a jockey, worked in circuses and nearly became a
church minister. He ultimately turned his life around as a result
of learning card tricks from a book of magic and spending long
hours practising them until he was a master card magician. He also
learnt tricks from others but turned those tricks into “masterpieces”
through his skill and creativity. Despite his label as being the master
of card tricks, ultimately his shows were massive in scale, requiring
truckloads of equipment for his elaborate illusions.
 Chapter 18

False Top Overhand Shuffle

Parameters and Generalisation

Conjuring
This shuffle leaves the top cards of the pack in place.

Computation
The procedures we have seen so far give a name to code that
does a single task. Parameters allow one procedure to do a whole
range of related things by providing the information that makes the
difference at the start. They allow a procedure to be generalised:
that is work in more general situations rather than just one
specific one.

The Trick
Pick up the deck as for a standard overhand shuffle. However, press the
bent second finger of the hand holding the cards up against the back of
the deck. Peel off a pile of cards from the front into the other hand as
in a normal shuffle. Make sure this first pile contains all the cards you
want to ultimately leave in place.

121
122 Conjuring with Computation

Continue with the shuffle, but as you drop the second group into
your other hand, pick back up the first cards at the back of the original
pack by gripping them between the thumb and other fingers.
You now have two groups of cards in your original hand: the remains
of the pack and, behind them, those you split off first with your second
finger between them. The new small pile you split off second is in your
other hand.
Push the tip of your second finger that is now between the original
last card and original first card, up against the original top card. With
this finger, you keep the former top cards in place at the back of the
pack in your original hand. Continue normally pulling piles of cards from
the pack in a standard overhand shuffle way. When you place the cards
from the bottom of the original pack into the other hand, you will be
left holding the group of cards that came from the top. Slide them down
at the front as though they had been at the back of the pack all along.

How It Works
The first cards that you peel off the front of the pack are held at the
back through the shuffle. They are put back at the front at the end.
It therefore guarantees that those cards are still on the top of the deck
at the end. By peeling off more than the number of cards you want
preserved in place, you can ensure that definitely that many cards are
left alone by the shuffle. You can do this shuffle as many times as you
wish, always peeling off a similar number to start.

The Magical Algorithm

This algorithm can be done repeatedly for more thorough shuffling.
TO DO a False Top Overhand Shuffle WITH number of cards to
remain in place n:
{n < 20}
1. Hold the pack of cards in one hand between thumb and fingers.
2. Push the second finger of that hand against the back (i.e., bottom
card) of the pack.
 False Top Overhand Shuffle: Parameters and Generalisation 123

3. Peel off a group of more than n cards from the front of that hand
dropping them into the other hand.
4. Move the rest of the pack in front of these, peeling off a second
group of cards from the front into the second hand, while pushing
the middle finger against the top card of the original group.
5. Lift off the rest of the pack with the original cards held at the
back, leaving only the second group behind.
6. DO UNTIL all cards are in the second hand apart from those being
held at the back.
a. Peel off a small packet of cards from the front of the original
deck into the front of the pile forming in the other hand.
7. Drop the remaining cards from the original hand to the front of
those in the other hand.
8. Pick up the deck in the original hand.
{The deck is shuffled, but the positions of at least the first n cards are
unchanged.}

Thinking Computationally
This procedure can be used in more varied situations than the previous
false shuffles. This shuffle doesn’t always do exactly the same thing.
Some tricks might need only the top card to be left in place. Another
might need the first dozen left alone.
We have provided a number n to the procedure that is used to guide
how many cards to leave at the front. The n can stand for different
actual numbers depending on what a particular trick needs. This is a
parameterised decomposition. We have created a description of a
general version of the shuffle where, when we use it in a trick, we specify
how many cards are to be left alone using the parameter, n.
In our algorithm, the parameter, n, is the number of cards to be left
in place. Notice we have a precondition that n is less than 20: (n < 20).
This is essentially saying that you should not rely on this shuffle to leave
close to half or more of the pack in place. If you try it, it may work, but
it probably won’t look so convincing. For it to be convincing, keep the
number of cards you preserve to be less than 20.
124 Conjuring with Computation

In an actual trick using this shuffle, we might, in an informal

description, say something like
“Do a False Top Overhand Shuffle, leaving 10 cards in place”.
The number 10 here is providing a value for parameter n on this
particular trick’s use of the shuffle.
In our slightly more formal notation, to mean the same thing we
would write:

(1) DO A False Top Overhand Shuffle WITH number of cards to

remain in place 10.

Here the number 10 is called an actual parameter, as it gives an actual

value to use. In different tricks (so uses of the procedure), this number
(so actual parameter) will be different. The variable n is called the
formal parameter. The formal parameter, n, is just a name used as
a placeholder for each and every value of n we might need. Of course,
for the trick to work, you will need to practise judging the right number
of cards to take. With a lot of practice, you may be able to do it exactly.
For most normal people, it will only be approximate, so best to err on
the side of caution and leave more cards alone than needed.
This idea of parameters is key in programming. Real programs are
built of lots of procedures each taking parameters in this way. Parameters
increase how generally useful a procedure is. For example, if we have a
procedure to print out the two-times table, we can change the 2 into a
formal parameter, n, in the code. Then, the procedure can print out any
times table, just by providing a different actual parameter.
This introduces the important computational thinking idea of
generalisation. Once we have written a program or procedure to solve
a specific problem, it is useful to generalise it: make it useful in as wide
a set of situations as possible. Adding parameters to procedures is one
way to do this.
When I first came across this false shuffle, I just saw it as a way to
leave the top card alone. That is useful, but the shuffle generalises to
leave a greater number of cards at the top of the deck. By generalising
that original shuffle as I understood it, it is now useful in more tricks.
 False Top Overhand Shuffle: Parameters and Generalisation 125

I do not have to invent new false shuffles for tricks that need more
cards left in place. Likewise, by generalising procedures, we save coding
work later.

PAUSE: Ada Lovelace

A countess, and daughter of poet Lord Byron, Ada Lovelace
(1815–1852) is often referred to as the first programmer because
of her work on algorithms for Charles Babbage’s proposed Victorian
computers. His Analytical Engine would have been the first general-
purpose computer if completed. She didn’t really write programs
though, not least because they had not developed a programming
language for Babbage’s machine. What she did do was work out the
sequences of states the machine would go through to follow specific
algorithms and, in doing so, found a mistake in an algorithm of
Babbage. Programmers today use the same idea to find bugs in
programs through logical thinking as part of code walkthroughs.
Perhaps Ada’s greatest achievement was to suggest that one day
computers would take over creative tasks, notably composing music,
when Babbage was envisioning the computer as just a souped-up
calculator to do arithmetic. She realised that as computers worked
with symbols, they could manipulate things other than just numbers.
This page intentionally left blank
 Chapter 19

Side Jogging Shuffle

Swapping One Procedure for Another

Conjuring
This false shuffle leaves the first quarter or so of the pack
unchanged.

Computation
If two procedures do the same thing, then you can swap one for
the other, and everything else should still work. This allows you to
make small changes without needing to change or check the rest
of the code.

The Trick
Hold the stacked deck in the overhand shuffle position. You want to
ensure the top of the stack (say the top 13 cards) remains intact. On
the first part of the overhand shuffle. pull off a large section of cards
from the front, more than the number you want to be left in place.
Place the next set of cards you pull from the front slightly off to one
side. Magicians call this side jogging. Shuffle the remaining cards in
small packets by doing a normal convincing overhand shuffle.

127
128 Conjuring with Computation

Figure 19.1. Side jogging involves lifting off the front section into the other hand
first, then putting the subsequent cards offset slightly. Finally, the original cards are
cut back to the front.

All that remains is to cut the cards, which you do by lifting the
section containing your protected card stack from the back of the pack.
It is easy because it’s side jogged from the rest. Move this section to
the top of the deck (Figure 19.1).
Practise until it looks like a convincing overhand shuffle and final
cut, while you have actually kept your stack in order on the top of the
deck.

The Magical Algorithm

TO DO a Side Jogging Shuffle WITH number of cards to remain
in place n:
{n < 20}
1. Hold the pack of cards in one hand between thumb and fingers.
2. Pull off a large packet of cards (more than n) from the front and
drop them into the other hand.
3. Pull off a second small packet of cards from the front of those
in the first hand, but drop them slightly to the side in the other
hand in front of the first packet of cards.
 Side Jogging Shuffle: Swapping One Procedure for Another 129

4. DO THE FOLLOWING UNTIL all cards are in the other hand.

a. Pull off a small packet of cards from the front of those in the
original hand with your thumb. Let them fall into the front of
the pile forming in the other hand.
5. Place the cards on the table.
6. Cut the pack of cards at the point where the cards are off to the
side.
7. Replace the old bottom part of the pack on to the top of the other
pile.
{The deck is shuffled, but the positions of at least the first n cards are
unchanged.}

How It Works
The side-jogged cards were the front cards, and they were replaced
unchanged back on the top by the final cut.

Thinking Computationally
The specification of this trick is identical to the previous false top
overhand shuffle. That means any trick that used that false shuffle could
swap in this one instead, and it would work just as well. You might do
that because you are doing two tricks in a show that start with the same
false shuffle. Therefore, it gives more variety to swap one for the other.
Or perhaps you are far better at one false shuffle as you have practiced
it more, so in an important show, you switch that shuffle into a trick
that is supposed to use the other. Or perhaps one is quicker, and you
have limited time to do the trick, or to do the cut, you need a table,
but you are doing the trick without one . . .
This is another important use of procedural abstraction. It gives
a framework for modifying algorithms, so tricks and programs, in a
way that is guaranteed not to affect, algorithmically, whether they work
or not.
With programs, we might want to swap in a new version of a
procedure that was faster or used less memory, perhaps, or one that
was just available in a new standard library that had been developed to
130 Conjuring with Computation

make the whole program easier to maintain (as we will see in the next
chapter).

Interlude: John Nevil Maskelyne

John Nevil Maskelyne (1839–1917) was one of the great Victorian-
era magicians. He invented, amongst many other tricks, the original
version of the levitation trick where a person appears to float above
the ground. His book Sharps and Flats (1894) gave a comprehensive
overview of the tricks of card sharps and how gamblers cheated at
card games. He created the Occult Committee, a group dedicated to
exposing frauds claiming to have supernatural powers. In response to
his debunking seances, spiritualist Alfred Russel Wallace (famous for
working out that evolution worked by natural selection independently
of Charles Darwin) just claimed that Maskelyne had supernatural
powers. Maskelyne also invented the mechanism behind the pay
toilet, where a penny was used to unlock a door, so is ultimately
responsible for the phrase “to spend a penny” meaning to go to the
toilet.
 Chapter 20

Cyclic False Cut

Libraries of Useful Procedures

Conjuring
This cut leaves the cyclic order of the pack unchanged, just starting
at a different point in the cycle.

Computation
Programmers collect together similar procedures such as ones
needed to manipulate a single data structure into what are called
libraries.

The Trick
Fan the cards and allow a spectator to touch the back of a card. Split
the pack at that point, moving the top half to the bottom. Repeat this
for as long as the spectator wants to, to “mix them thoroughly”.

131
132 Conjuring with Computation

The Magical Algorithm

TO DO a Cyclic False Cut:
{Nothing is assumed.}
1. DO THE FOLLOWING UNTIL the spectator is happy the cards
are thoroughly mixed.
a. Fan the cards.
b. Ask a spectator to point to a card.
c. Split the pack at that point, placing the top part underneath.
{The deck appears shuffled, but the cyclic order of the cards is
unchanged, though starting at a different point in the cycle.}

How It Works
This is actually just a series of normal cuts of the pack. Cutting the
pack does not change the cyclic order of the cards. You can do it as
many times as you like, and the cycle will be the same, just starting
at a different point. To see this, imagine the cards laid out in a circle,
before the cut and then again after the cut. The circle is the same before
and after the cut, just rotated (see Figure 20.1). The same cards follow
each other as before. You can do it as many times as you wish, and the
cycle remains. By fanning the deck each time, it is less obvious what is
happening than just cutting the deck repeatedly. You can also let lots of
people take part, so making it a communal decision as to how the pack
is “shuffled”.

Thinking Computationally
The techniques from this section are useful components that can form
part of a range of tricks. Any good magician will build up a repertoire
of them, to call on as needed when inventing new tricks or developing
their own versions of old ones. A false shuffle at the start of a trick, that
relies on cards being in a known position or order at the start, will throw
the audience off the trail of what is actually happening.
In programming, the idea of a library is similar to this. Calling it a
“library” is actually a design metaphor using the idea of a library of
 Cyclic False Cut: Libraries of Useful Procedures 133

Figure 20.1. The cards of the pack laid out in a circle before and after a cut. The
cyclic order is unchanged.

books to make people think of an organised collection. Here we collect

together useful similar magical techniques: a magician’s repertoire of
cuts and shuffles to create a “Cuts and Shuffles” library. In programming,
we have libraries of useful operations that can be performed on one data
134 Conjuring with Computation

structure, like a list manipulation library, that gives you everything you
might need to manipulate lists of things, or a String library of operations
for manipulating strings of characters (words, sentences, etc.).
Rather than describing all the procedures in detail whenever they
are needed, giving the full algorithm of every trick or program, we can
just include the library noting they are available. If writing a series of
magic manuals to record all our tricks, we might similarly write one book
just about cuts and shuffles, and refer to it whenever we mention one.
We would then avoid having to explain each, so avoid writing out the
instructions, like those for a cyclic false cut, in every book.
This is an organising principle that allows us to make good use of
procedural abstraction and decomposition, reducing the amount of
work and actually making our instructions clearer.

PAUSE: Maurice Wilkes and

David Wheeler
Maurice Wilkes (1913–2010) and his team developed one of the
first digital computers, EDSAC, at Cambridge University. It was
immediately used by researchers. Three Nobel Prize-winning teams,
in Chemistry, Physics and Medicine, used EDSAC in their research.
It hosted the world’s first computer game: a noughts and crosses
program. With EDSAC team member, David Wheeler (1927–2004),
Wilkes created a library of useful procedures that could be included in
any program. They were stored on paper tape in a filing cabinet, but
descriptions of them (their specification or interface) were stored in a
library catalogue. It contained the details of what each procedure was
for as well as how to call them. Today we would call this catalogue
an Application Programming Interface (or API) specification.
 Part 6

Building Bigger

We have now seen most of the basic techniques that allow us to build
large programs a little at a time. In this section, we first look at the
importance of clear descriptions of what procedures do, then look at
how to chain results from one procedure or trick to the next and apply
the techniques to a trick we have already seen.
This page intentionally left blank
 Chapter 21

The False Choice

Programming Interfaces

Conjuring
The false choice is a way of leaving a volunteer thinking they chose
one of three things when actually you did.

Computation
Procedures need to have clear interfaces, i.e., specifications or
descriptions of exactly what they do and when they can be used if
they are to be used in large programs without introducing mistakes.

The Trick
Suppose there are three objects, and you wish to be left with a specific
one, say the third. This could be three piles of cards, for example, or
three boxes with a reveal hidden in one of them. Ask the volunteer to
point to one of the objects. If they point to the one you wish to keep,
then say “Ok, we will keep this one”. You remove the other two objects.
If, on the other hand, they point to one of the ones that you wish
to lose, then say “Ok, we will remove that one”. You remove it. You
still have, in this second case, another object to lose. You now fill up

137
138 Conjuring with Computation

some time with a story that fits the situation, so everyone loses track
of what you did originally. Then you ask them to point to one of the
two remaining objects. Whichever they point to, you remove the one
you wanted to lose, and keep the other, saying the appropriate thing to
justify this, as above (see Figure 21.1).

Figure 21.1. If (a) the person points to the object you intend, then you take that
as the choice. Otherwise (b, c) you note the pointed-to object is discarded and do a
further false choice.
 The False Choice: Programming Interfaces 139

The Magical Algorithm

TO DO a False Choice WITH object n chosen:
{Three objects are laid out. You wish object n of the three to be chosen
where n is either 1, 2 or 3.}
1. Ask a volunteer to point to one of the objects.
2. IF they point to object n
THEN
a. Say “OK. You have chosen that one.”
b. Discard the other two.
OTHERWISE
a. Say “OK. We get rid of that one.”
b. Discard the object pointed to.
c. Talk about something related to give time for them to forget
exactly what you did.
d. Ask the volunteer to point to one of the remaining two objects.
e. IF they point to object n
THEN
i. Say “OK. You have chosen that one.”
ii. Discard the other one.
OTHERWISE
i. Say “OK. We get rid of that one.”
ii. Discard the object pointed to.
iii. Say “We are left with this one, pointing to the remaining
object, n.”
{Object n is left.}
140 Conjuring with Computation

How It Works
Make sure you say “point” not “choose” when asking them to pick a
pile as that is neutral language that does not imply anything about
what you will do once they point to it. Because you did not tell the
person what you were going to do with the object they pointed to and
used such neutral language they see nothing strange over whether you
remove it or keep it. As long as you only do the trick once they have no
idea, there was some alternative course of action that could have been
taken.

Thinking Computationally
It is not just cuts and shuffles that can be used to enhance different
tricks. Here we have given a generally useful false choice. In a similar
way, there are many different kinds of procedures that can be written
and stored in libraries to be pulled out when needed. Here we have a way
of forcing a particular choice while making it look like a free choice. “The
Forcing Matrix”, described in the next chapter, is another example. We
could therefore have a separate collection, or “library”, of false choice
algorithms too.
Each separate mini trick like this, or library procedure, we create
needs a clear interface in the sense of it being designed, so there is a
clear situation when it can be used having a clear and precise effect.
This then makes it easy to give a precise description (i.e., specification)
of when it can be used and what the effect is. For example,

The False Choice can be used when you have three objects (or
piles of objects) and wish to be left with just one specific one of
the three.
After the false choice, you are left with the one of the three
you wanted, but the audience believe that an audience member
made the choice.
 The False Choice: Programming Interfaces 141

We have tried to make sure all the cuts, shuffles and false choices
described here do have clear interfaces. For example, an interface for the
Cyclic False Cut (Chapter 20) is

The Cyclic False Cut is a false shuffle that can be done any
time that the precise order of the cards does not matter, only
the cyclic order matters. It will leave the cyclic order alone, just
change the starting position.

Specifications like this are often written as comments (notes for a

human reader) before the actual code in programs.
Using English to do this can sometimes be a problem though, as
it is hard to be very precise. This leads to the idea of logic-based
specification languages where maths is used to give the specification.
It would specify precisely, using maths, that a number of cards should be
less than 20, for example. We have essentially done that in writing some
of the simpler preconditions like {n < 20} in our earlier false shuffle
algorithms. We could similarly write specifications such as “the cyclic
order is unchanged” using logic. Writing mathematical specifications is
very like writing programs, as we are using a fixed and precise notation
to say what we want to say. The difference is that, rather than saying
how something is done as we do in a program, we just describe the end
result: what should be the case, what should have happened. Ideally,
a combination of both English and logic would be used, so it is both
precise (the logic) and easy to understand in overview (the English).
Writing mathematical specifications to describe program interfaces is
really important when the code is safety-critical. We won’t go into the
details here of how to write fully mathematical and logical specifications,
as it would take a book on its own.
142 Conjuring with Computation

PAUSE: Tommy Flowers

Tommy Flowers (1905–1998) was an unsung hero of the invention of
the computer. He worked for the British General Post Office working
on electronic telephone exchanges. He was asked to help Alan Turing
and others at Bletchley Park who were trying to make machines to
help quickly crack the German ciphers. In particular, he helped build
machines to crack messages sent by the German high command
using a cipher that was even more complex than the famous Enigma.
Flowers came up with the design of the Colossus: the world’s first
programmable digital computer actually built. It used the kind of
valves then used in telephone systems. Others in the team were
sceptical, so he worked on the ideas on his own, back at the Post
Office, using his own money. His Colossus machines were eventually
used at Bletchley Park and cracked one of the most important
messages of the war. It showed Hitler didn’t believe the Normandy
landings would happen, so didn’t add reinforcements. Knowing
this, General Eisenhower ordered D-day to go ahead the next day,
changing the course of the war. Because work at Bletchley Park was
kept secret after the war, no one knew of Flowers’ contribution for
30 years.
 Chapter 22

The Forcing Matrix

Functions

Conjuring
The forcing matrix gives a way to always force the selection of the
same number, despite a volunteer having a series of free choices,
choosing numbers from a grid to add together.

Computation
We have previously used decomposition to create procedures which
just get a job done. A variation is the function. The job of
a function is to deliver an answer, such as a number or data
structure. That answer is used later in the program. We also look
at preprocessing data: organising data in advance, with the result
used over and over again later in the program.

The Trick
You set out a grid of “random” numbers. There are various ways you
can do this such as dealing the top cards from a pack of cards into

143
144 Conjuring with Computation

a 4 × 4 grid, writing out numbers 1–25 in a 5 × 5 grid (1, 2, 3, 4, 5, . . .),

or just having a prepared card containing “random” numbers.
Different members of the audience take it in turns to pick a number
on the grid. They circle the number chosen. They then cross out the
other numbers in the same row and same column as their chosen number.
You now have a series of randomly picked numbers in the ones
circled. You add them up to get a final random number. In doing this
as a trick in its own right, you reveal that the final total is the number
you previously sealed in an envelope (or reveal as a prediction in some
other creative way of your choosing). Alternatively, it can be a forced
number used as the basis of some other main trick (such as in the next
chapter).

The Magical Algorithm (Part 1)

TO DO Use the Forcing Matrix WITH forcing matrix f :
1. DO THE FOLLOWING UNTIL all numbers in the matrix, f, are
circled or crossed out.
a. Ask an audience member to pick and circle a number that has
not already been circled or crossed out.
b. Cross out all other numbers in the same row as the circled
number.
c. Cross out all other numbers in the same column as the circled
number.
2. Add the circled numbers to get a total, n.
3. RETURN the number n.

How It Works
To set the trick up, you first need to decide the number you wish to
force and the size of grid you wish to use. Let’s say the number you
want to force is 127 and you want to use a 5 × 5 grid. The grid has
5 rows and 5 columns. You need a “seed” number for each row and
each column, so in this case, you need 5 + 5, i.e., 10, seeds. Create any
 The Forcing Matrix: Functions 145

(a) (b) (c)

Figure 22.1. Creating a forcing matrix: (a) seed numbers labelling rows and columns
that add up to the number you want to force, 127. (b) Fill in the numbers by adding
row and column seeds. (c) The final forcing matrix is produced by removing the seed
row and column.

10 seed numbers that add up to the forcing number (e.g., 0 + 48 + 12 +

3 + 7 + 13 + 1 + 2 + 37 + 4 = 127). Write the first 5 seed numbers down
the side of a 5 × 5 grid and the other 5 numbers along the top. Now,
for each position in the actual grid, add the seed at the start of that
row to the seed at the top of the column, and write the answer in that
cell. So with the above seed numbers, you would get the table shown in
Figure 22.1.
Removing the seed numbers gives the final forcing matrix to use in
the trick to force the number 127.
You now do the trick above using this grid, and the number that
results will be 127. If you start with a different initial forcing number,
and seeds that add up to it, you can use the trick to force any number
you wish.
The forcing number linked to a grid just with numbers 1–25 is 65.
Try creating it yourself using seeds: 0, 5, 10, 15, 20, 1, 2, 3, 4, 5. If you
did the trick with the resulting forcing matrix, you would always get the
number, 65.
Similarly, to force the number 20, you could use a 4 × 4 grid with
seeds: 1, 2, 3, 4, 1, 2, 3, 4. Again try it yourself to see what forcing matrix
you end up with. As all these numbers are less than 10, you could create
this matrix by dealing out playing cards into a grid (having placed the
right cards in the right order at the top of the pack) and even have
146 Conjuring with Computation

done a false shuffle such as a false top overhand shuffle (Chapter 18) to
appear to have shuffled the pack while leaving those cards in place. Use
your imagination to invent more creative ways of making the grid!
Why does it work? Each position is the total of a row and column
seed. When you cross out everything else in that row and that column,
you cross out all other chances of picking those two seeds. That means
each choice gives you two of the seeds that have so far not been used
to add in to the total. Eventually, when all numbers are chosen, you
have all the seeds represented in the numbers chosen and so part of the
addition.

The Magical Algorithm (Part 2)

TO DO Create a Forcing Matrix WITH size n, forcing number f :
1. Choose 2 × n numbers that add up to total, f.
2. Write the first n of these numbers down the side and the other n
along the top of an n by n grid. These are the row and column
seed numbers.
3. FOR EACH position in the grid
a. Add the row seed number to the column seed number.
b. Write the total in that position in the grid.
4. Write out a neat copy of the grid without the seed numbers. This
is your forcing matrix.
5. RETURN the resulting forcing matrix.

Thinking Computationally
We have up to now been decomposing magic tricks (and programs) into
procedures whose purpose is to get a job done. Here we are seeing a
variation of this. Instead of just getting a job done, we are decomposing
into parts whose purpose is to return a result (like a number calculated,
 The Forcing Matrix: Functions 147

or a grid produced). We call these functions rather than procedures, as

the purpose is often to compute a mathematical function. Both functions
and procedures are sometimes more generally called subroutines (and
all three terms are often interchanged and in some situations also called
methods just to be even more confusing).
Here we have two functions. We have a function whose job is to
create, so return, a forcing matrix. We then, as part of an actual magic
trick, use the forcing matrix to give us (so return) a number. That
forced number is used later in the next part of the trick.
Something else important is going on here. It is the idea of
preprocessing data. Magicians do something like this often. They
prepare some prop, like the forcing matrix, beforehand. They only need
to do this once, but then can use it over and over again, each time they
wish to do the trick.
The equivalent thing programmers do is to write code that prepares
a data structure that helps do some tasks quickly. Preparing a data
structure to make fast searching possible is a good example. We search
the same data over and over again, so it is worth spending some time
preparing the data to make those searches fast. That is what search
engine companies do. When you type a search query into a search engine,
the search engine is not actually visiting and checking every page out
there on the web directly in that fraction of a second. It is checking a
pre-prepared data structure that is held by the search engine company;
based on them, having preprocessed the web pages, their software bots
have found trawling the Internet.
This is exactly the same general idea as an old-fashioned paper
telephone directory (or an index in a book like this one). Someone spent
time organising (in this case, sorting) the data into alphabetical order
(to create a sorted list data structure) once. That work is done once
so that the many later searches for telephone numbers (or pages) can
be done quickly later.
148 Conjuring with Computation

PAUSE: John McCarthy

John McCarthy (1927–2011) was one of the founders of Artificial
Intelligence and behind the first functional programming language,
LISP. It was the language of choice for early programmers working
on Artificial Intelligence, trying to write intelligent programs. The
idea is that programs are completely built out of mathematical
functions. This makes them very mathematically pure and easy to
reason about. It turned the ideas of mathematician, Lorenzo Church,
who developed a logic based on functions, into a programming
language. His logic demonstrated the limits of computation in a
similar way to Alan Turing’s “Turing Machines”: what a computer
could and could not do. He was a strong supporter of free speech
and argued that computers could have beliefs. He wrote a satirical
short story The Robot and the Baby about a future society where
robots were banned from being human-like or being programmed to
have emotions. It predicted the trend for instant social media-based
news leading to millions posting their views on news as it happens.
 Chapter 23

Forced Wizardly Book Magic

Calling Functions — Building Bigger
and Better

Conjuring
A number is randomly chosen. An audience member turns to that
page in a book. The selected page then magically directs them to
a word. It is both on a random page and a random word on that
page: a word no one could have possibly predicted. . . but you did.

Computation
Once you have decomposed a program into functions and written
the functions, those functions provide answers to be passed on to
the next procedure or function in a sequence.

The Trick
This trick just combines the forcing matrix trick with Wizardly Book
Magic (Chapter 6) to add an extra level of mystery.

149
150 Conjuring with Computation

Before the trick, find an interesting page (say page 127) in your
chosen book (perhaps in The Wonderful Wizard of Oz) and work out
the predicted word on that page as in Wizardly Book Magic of Chapter
6. Then create the appropriate forcing matrix (say of size 5 × 5) for that
page number, i.e., that forces 127. Write the predicted word from that
page in an envelope ready for the reveal at the end.
When doing the trick, use the forcing matrix to pick a page in the
book apparently at random, but actually known to you in advance. Then
use that page to do the Wizardly Book Magic trick. It finally gives you
a word in the book that is your final prediction. Have someone open the
envelope to reveal that you predicted the word.

The Magical Algorithm

TO DO Forced Wizardly Book Magic WITH book b, forcing
matrix size n, forcing number p1:
1. LET f BE THE VALUE RETURNED BY
DOING Create a Forcing Matrix WITH size n, forcing number p1.

2. LET p2 BE THE VALUE RETURNED BY

DOING Using the Forcing Matrix WITH forcing matrix f.

3. DO Wizardly Book Magic WITH book b, page p2.

Note as the forcing matrix returns a number we know in advance, we only
do the second step for presentation purposes, not really to generate a
new number: p1 and p2 are the same numbers. We have assumed here
that we have first written a parameterised version of Wizardly Book
Magic set up for a given page in a given book.

How It Works
This is combining two tricks that each is guaranteed to force a result,
with one leading to the other. If you understand how each works (see
Chapters 6 and 22), then it is clear the combined trick is guaranteed to
work.
 Forced Wizardly Book Magic 151

Thinking Computationally
This involved combining two tricks, by passing the result of one to the
next, to give a bigger, better trick. Programmers do the same thing,
chaining together existing functions that do separate things. The first
provides the input to the next. Programs are often decomposed in this
way, using a similar structure to the above algorithm. A first function
returns a value or data structure that it computes (like the matrix).
That value is passed to the next function as a parameter. The second
function, in its turn, uses the value and calculates from it something new
(like the page number). It returns that to be used in the next function,
and so on.
In the above, forcing matrix, f, is the result of the function that
creates the forcing matrix. It is then the actual parameter (i.e., value
passed) to the function that gives instructions on how to use a forcing
matrix. A similar thing then happens with the page number. It is the
result returned by using the forcing matrix, and then that page number
value is passed to the actual Wizardly Book Magic trick.
This also shows that to create a better trick, you don’t necessarily
have to combine just simple trick elements like shuffles. You can
sometimes combine whole tricks, that are magical on their own, to get
much more powerful tricks.
A similar thing is done with programs, building ever more sophis-
ticated programs, by combining functions that already do something
useful and have been used as a self-contained program so that the
results of one are passed to another. Procedures are also used when
an action is needed on the results calculated so far by the chain of
functions. For example, a procedure might be used to display the results
of the function on the screen. By building up ever large sequences, more
complex programs are created.
152 Conjuring with Computation

Interlude: David Devant

Claimed by some to be the greatest magician of the 20th century,
David Devant (1868–1941), was famous for the humour in his
magic tricks, ensuring that they were far more entertaining than just
the illusion itself. First President of the Magic Circle, he stressed
the importance of performance, and once put an arrogant young
magician in his place who was boasting about how many tricks he
knew, by saying he only knew a few tricks but what mattered was
that he could do them very well. He was once accosted in the street
by a patient from a psychiatric hospital who demanded he make
coins appear out of thin air. Devant kept him entertained by doing
so until people from the hospital came to take the patient back. He
was passionate about new technology and when cinema was invented
became one of the first people to tour the country showing films.
Several films were also made of him performing magic including
pulling rabbits out of a top hat.
 Chapter 24

Invisible Palming (Again)

Sequencing Procedures Together

Conjuring
We return to Invisible Palming and the Teleporting Top Hat and
look at them anew.

Computation
Procedures with parameters allow us to create generalised pro-
grams. We can decompose them into parameterised parts and
sequence them one after the other, using the same parameters.
By matching preconditions and postconditions, we can give an
outline argument as to why the whole algorithm works without
worrying about how the parts work.

The Trick
In this chapter, we are going to just look again at the same Invisible
Palming trick from Chapter 1 and its variation from Chapter 14, the
Teleporting Top Hat. In doing so, we will put together some of the ideas
we have seen so far.

153
154 Conjuring with Computation

Thinking Computationally
We wrote out the original Invisible Palming trick as a single algorithm
that wasn’t decomposed at all. We then saw in the Teleporting Top Hat
that with a little decomposition we could make it easy to substitute a
new part with an identical effect. Here, we are going to see how we
could decompose it further into a whole series of procedures, but also
use parameters to make it more general. We also use specifications to
make clear the intended effect of each part. In doing so, we are seeing
the foundations of the way large programs are built.

Generalising the trick

First of all, how can we make the trick more general? The number of
cards used in the trick could be a parameter, as it is something that
could in theory change. Our main command (a procedure call) to do the
trick is then:
DO Invisible Palming WITH 15 cards.
So our definition of the trick now takes a parameter:
TO DO Invisible Palming WITH n cards: . . .
Does it work for any number of cards? See if you can work out what
numbers of cards it does work for before reading on.
Fairly obviously, it does not work for all numbers as there must be
an extra card to appear to have moved from one pile to another. That
means the number of cards, which we are calling n, has to be odd. It
would not work with 14 cards or with 16 cards. We can make this clear
with an assertion to use as a precondition:
{There are n cards AND n is odd.}
That on its own is not enough though. It doesn’t actually work for
any odd number of cards. When you split the cards into two piles, before
the last card is placed, you need two identical piles containing a series of
pairs but with one extra in each (so both must contain an odd number
of cards at this point). Then, you want to place one extra card in one
of the two piles. Fifteen cards work because it splits into 7 cards and
 Invisible Palming (Again): Sequencing Procedures Together 155

7 cards (both odd) with that last card making 8 cards in one pile. In
each final pile, there are 3 pairs, plus one extra in each to make 7 and
then that last “extra card” to put on one of the two. As you have two
odd piles, that “extra card” disappears into a new pair whichever pile it
is placed on.
Thirteen cards doesn’t work because it splits into 6 cards and 7
cards: 3 pairs in each pile and one leftover. The 6 pairs form two piles of
6, and now whichever pile you place the “extra card” on cannot complete
a pair, so doesn’t appear to disappear.
When we take away the extra card (so have n − 1 cards), then we
need two piles of an odd number of cards. So, that number (n − 1) of
cards must divide by 2, and when halved the resulting number of cards
should be odd. This gives us a new fact about n that:

{(n − 1) ÷ 2 is odd.}

Combining this with the fact that n itself also has to be odd, we get
the precondition for the trick:
{There are n cards AND n is odd AND (n − 1) ÷ 2 is odd.}
This says that the trick is only guaranteed to work if these three
assertions are true. We could treat it as an instruction to execute that
essentially terminates our attempt to do the trick if we find we have the
wrong number of cards. If that happens, we need to start again with the
right number. I always count the cards to double-check this precondition
holds before doing this trick, exactly for that reason.
Our top-level instructions in our new written version now just give
an overview of the trick. By the end, a card should have appeared to
jump piles.
TO DO Invisible Palming WITH n cards:
{There are n cards AND n is odd AND (n − 1) ÷ 2 is odd.}
1. DO Deal out the cards in pairs with one left over WITH n cards.
2. DO Take the cards back as pairs splitting the pairs into two piles
WITH n cards.
3. DO Place the extra card.
4. DO Pretend to move a card between piles.
156 Conjuring with Computation

5. DO Reveal that the extra card has magically moved piles WITH
n cards.
{An extra card appears to have moved piles.}
We can still get the detail of exactly what we need to do when we
want it by looking at the separate instructions for each step. However,
this gives an overview, uncluttered by all the detail.

Chaining assertions
We still need to write out the detailed instructions for each step as its
own procedure. Each can have its own precondition and postcondition,
saying what it does. Actually writing out those procedures’ assertions
can help us check overall the trick is going to work BEFORE we worry
about the detail. In the sequence of instructions, each postcondition
should match the precondition of the next step. If it does not, we might
be using an invalid step. The first procedure call has to guarantee the
precondition of the second, and the second has to do the same for the
third. This provides a chain of reasoning as seen in Figure 24.1.
The first step in the algorithm is to deal out the pairs with one left
over. It has the same precondition as the overall algorithm: the same
thing must be true at the start for both. The first step guarantees that
on its completion (its postcondition):
{There are an odd number of pairs of cards dealt out with one extra
card on its own, totalling n cards.}
That is exactly the precondition of the next procedure to be followed.
Our preconditions and postconditions therefore are giving us an outline
logical argument as to why the trick as a whole works, step by step. The
argument runs something like this:

If we start with the right number of cards to make two odd piles with
one extra card (i.e., n is odd AND (n − 1) ÷ 2 is odd . . .
then on dealing them out into pairs . . .
There are an odd number of pairs of cards dealt out with one extra
card on its own, totalling n cards.
After collecting them back in, there are two piles each holding . . .
and so on.
 Invisible Palming (Again): Sequencing Procedures Together 157

Figure 24.1. Linking the procedures together with matching preconditions and
postconditions.
158 Conjuring with Computation

Writing out the instructions for each step

as procedures
Now, we have a clear specification for each step. If each step meets its
specification, then we know from our logical argument that the whole
trick works. Now we write out the detail of each of those major steps,
writing the definition of the procedures, separately making sure each
does guarantee its postcondition if starting with the precondition being
true.
The separate steps are just groups of instructions based on those of
our original version of the algorithm, except with parameters, using n
instead of the fixed number 15 as in the original. This means the number
of times we, for example, place cards is also expressed in terms of n:
DO THE FOLLOWING (n − 1) ÷ 2 TIMES . . .
where (n − 1) ÷ 2 is just the number of pairs we have if we have n cards
in total.

In summary
Overall, we have written the instructions for a general version of the
trick that works with any appropriate number of cards as specified. We
have decomposed it into a series of major steps and used preconditions
and postconditions for those steps to give an outline argument as to
why they work. Finally, we have written the detail of those decomposed
instructions. If we can give a convincing argument that each meets its
own specification, then our overall argument shows that the whole trick
works.

The Magical Algorithm

DO Invisible Palming with 15 cards.
TO DO Invisible Palming WITH n cards:
{There are n cards AND n is odd AND (n − 1) ÷ 2 is odd.}
1. DO Deal out the cards in pairs with one left over WITH n cards.
2. DO Take the cards back as pairs splitting the pairs into two piles
WITH n cards.
 Invisible Palming (Again): Sequencing Procedures Together 159

3. DO Place the extra card.

4. DO Pretend to move a card between piles.
5. DO Reveal that the extra card has magically moved piles.
{An extra card appears to have moved piles.}

TO DO deal out the cards in pairs with one left over WITH n
cards:
{There are n cards AND n is odd AND (n − 1) ÷ 2 is odd.}
1. DO THE FOLLOWING (n − 1) ÷ 2 TIMES.
a. Place 2 cards in a pair.
b. Say “Two cards make a pair”.
2. Place a single card in a pile on its own.
{There are an odd number of pairs of cards dealt out with one extra
card on its own, totalling n cards.}

TO DO take the cards back as pairs splitting the pairs into two
piles WITH n cards:
{There are an odd number of pairs of cards dealt out with one extra
card on its own, totalling n cards.}
1. DO THE FOLLOWING (n − 1) ÷ 2 TIMES.
a. Take a pair and split them into the two piles being formed.
b. Say “Two cards make a pair”.
{There are 2 piles each holding (n − 1) ÷ 2 cards, where (n − 1) ÷ 2 is
odd, with one extra card still to be placed, totalling n cards.}

TO DO place the extra card:

{There are 2 piles each holding (n − 1) ÷ 2 cards, where (n − 1) ÷ 2 is
odd, with one extra card still to be placed, totalling n cards.}
1. Have a volunteer choose one of the two piles to put the extra
card.
2. They place the card on that pile.
{There are 2 piles of cards, one with an even number of cards, one
containing an odd number of cards with one extra card.}
160 Conjuring with Computation

TO DO pretend to move a card between piles:

{There are 2 piles of cards, one with an even number of cards, one
containing an odd number of cards with one extra card.}
1. Place a hand over the pile with the extra card.
2. Rub the back of that hand.
3. Lift your hand, showing the palm before placing it over the other
pile.
4. Tap your hand and remove it, saying the card has moved.
{There are 2 piles of cards, one with an even number of cards, one
containing an odd number of cards with one extra card.}

TO DO reveal that the extra card has magically moved piles:

{There are 2 piles of cards, one with an even number of cards, one
containing an odd number of cards with one extra card.}
1. Pick up the first pile (where the “extra” card was placed).
2. DO THE FOLLOWING UNTIL you are holding no cards.
a. Add 2 cards from the pile you are holding, to a pile on the
table.
b. Say “Two cards make a pair”.
3. Pick up the second pile (where the “extra” card was NOT placed).
4. DO THE FOLLOWING UNTIL you are holding one card.
a. Add 2 cards from the pile you are holding, to a pile on the
table.
b. Say “Two cards make a pair”.
5. Reveal that an extra card is left in your hand. The extra card was
in the second pile.
{A card appears to have moved piles.}
 Invisible Palming (Again): Sequencing Procedures Together 161

PAUSE: Robert W. Floyd and

Tony Hoare
Robert W. Floyd (1936–2001) originally studied liberal arts at
university, and Tony Hoare studied classics and philosophy. Both
went on to become great computer scientists. Floyd’s biggest
contribution was in the idea of using assertions attached to points
in flowchart versions of programs as a way of proving that they
did the right thing. These ideas were built on by Tony Hoare, who
created a programming logic: a set of rules that could be used to
reason about exactly what a program did and so whether it met
a mathematical statement of what it should do (its specification).
Between them, they kick-started the field of automated program
verification, vital for programs that are safety-critical: where bugs
can kill. Both made many other contributions to computer science
during their lifetimes from inventing really important algorithms to
the design of programming languages.
This page intentionally left blank
 Part 7

Abstraction and Data Representation

We’ve seen abstraction applied to the way we write out the order of
instructions in the form of procedures. Now, we explore similar ideas
with data: data abstraction. Data can have different representations
which are abstract versions of how the data is really stored. We look at
what it means to hide the underlying representation of data.
This page intentionally left blank
 Chapter 25

Drawn to You
Codes Representing Data

Conjuring
Four people each draw a picture on a piece of paper. The paper
is jumbled up, but you are able to read each person’s personality
and work out who drew each.

Computation
Tricks often rely on representing information in some particular
way. Choosing an appropriate data representation is core to
computational thinking and programming.

The Trick
You tear the bottom half of a sheet of paper out of an A4 pad. Fold this
piece in half, then half again and tear it into quarters along the folds.
Hand a piece to each of four friends who you have asked to stand in a
line. Next, give each an identical pencil and ask them to secretly draw
a simple picture on their piece of paper. You may want to give them

165
166 Conjuring with Computation

a book to lean on while doing this. Each can draw anything they like.
Have each crumple their paper into a ball and put it into a small bag
where they are well and truly mixed.
You announce that you are able to judge the personalities of
people from their drawings. You can also read personalities from their
body language (which you have been observing). You will now do a
psychological reading of their picture and match each to the person
who drew it.
You remove the screwed-up bits of paper one at a time, uncrumple
them, and through the psychological reading of the picture you correctly
identify which person drew which picture!

How It Works
Though it seems that simply tearing the half sheet into four is an
innocent action, it’s actually the secret. It encodes information on the
bits of paper.
If you look at four bits of paper torn from a pad in this way, you will
see that each is similar, but actually subtly different (see Figure 25.1).

• The bottom right quarter has two smooth, clean edges, the
original corner of the sheet from the pad you started with. It
is the only piece that has two such smooth, clean edges.
• The bottom left section has only one smooth edge (from the
bottom edge of the original sheet), but that edge is long (about
10 cm if you used A4 paper).
• The top right section also has only one smooth edge (from the
side of the original piece of paper), but it is a shorter side (roughly
7 cm if you used an A4 pad).
• The top left quarter has no smooth edges at all. It is torn all
around, as it came from the middle of the original piece of paper.

By giving out the paper squares in the above order, each person has been
given an identifiable piece of paper. Instead of looking at the drawings,
you can just look for the different pieces of paper.
 Drawn to You: Codes Representing Data 167

Figure 25.1. The bottom half of a piece of paper torn out of a pad of paper. Each
side has different smooth sides.

The Magical Algorithm

TO DO Drawn to you:
1. Ask four people to stand in a line, giving each an identical pencil.
2. Tear the bottom half of a sheet of paper out of an A4 pad.
3. Fold this piece in half, then half again and tear it into quarters
along the folds.
4. Hand a piece to each person, in the order they were torn off:
Person 1 — bottom right quarter; Person 2 — bottom left quarter;
Person 3 — top right quarter; Person 4 — top left quarter.
5. Ask them to secretly draw a simple picture of their choice on their
piece of paper.
6. Have them crumple their paper into a ball and put them into a
small bag.
7. Mix up the balls of paper in the bag.
168 Conjuring with Computation

8. Announce that you are able to judge the personalities of people

from their drawings.
9. DO THE FOLLOWING 4 TIMES.
a. Pull out a drawing.
b. Un-crumple it.
c. IF the paper has two smooth edges
THEN Announce that it was Person 1.
IF the paper has one long smooth edge
THEN Announce that it was Person 2.
IF the paper has one short smooth edge
THEN Announce that it was Person 3.
IF the paper has no smooth edges
THEN Announce that it was Person 4.

Thinking Computationally
A big part of computational thinking involves working out the best way
to represent information to make the task at hand easy. Western
cultures use characters like A and 1 strung together into words and
numbers to represent information, but there are lots of different ways
to represent the same information. In many oriental cultures, characters
are used that represent words. In South America, knotted strings, called
quipu, were once used to represent numbers and possibly words . . .
Choosing a good representation matters. This trick is based on
choosing a representation for four different things (the four people,
or equivalently the four numbers representing their positions). The
representation is one that you, the magician, can easily see, but they do
not notice at all. The audience just see the four bits of paper as the same.
However, the performer, in the way they tear them out, is building their
own secret representation of the vital information. For them, knowing
the representation is just like writing on the corner: Person 1, Person 2,
Person 3 and Person 4.
To help remember, you could even go a step further and create
a representation in words: a code, where S represents a smooth side:
(1) S-S, (2) LONG-S, (3) SHORT-S and (4) NO-S. Remember the
code in that order, and it is even easier to remember which is which.
 Drawn to You: Codes Representing Data 169

This is made easier still if you organise the people into a line rather
than just letting them stand anywhere. By doing that, you physically
give the people an order too (their position is now a representation of a
number). Their order matches the order you have remembered for the
paper. You are organising the information they embody by positioning
them in the line. The organisation of information (its structure and
representation) makes the task easier to do. Organising information into
the most convenient form is what data representation is all about.
Programmers often choose how to organise data before writing code.
For example, if you sort names into an alphabetically ordered list first,
then it is possible to find a given name much faster. You can use fast
search algorithms. That is why dictionaries and the index of a book are
in alphabetical order, as were old-fashioned telephone directories. Search
engines organise the information from web pages before you do searches,
allowing them to search every web page in the world and return a list of
those that are relevant in a fraction of a second.
Choose the right way to organise information for the task, and that
task can be easy or hard, whether it is a computer or a person doing the
computation.

The data representation

Here we represent people as numbers and as properties of pieces of
paper, as well as then ultimately remembered code words. There are
thus four different representations of the same thing (see Figure 25.2).
Actually, there is a fifth too in the drawings, which it is asserted
represents the personality of each person, but that is actually ignored.

Person Position Physical Properties Code

Number of Paper
First person 1 Two smooth sides (S-S)
Second person 2 One long smooth side (LONG-S)
Third person 3 One short smooth side (SHORT-S)
Fourth person 4 No smooth sides (NO-S)

Figure 25.2. The way the information about the people is represented in the paper
for the “Drawn to You” trick.
170 Conjuring with Computation

Interlude: Joseph Jastrow

The type of mentalism effect where people claim to be able to
read personality through objects is called psychometry. Psychometry
was hyped by Joseph Buchanan in the mid-1800s as revolutionary
science. Experimental psychologist Joseph Jastrow (1863–1944),
who made a career of using scientific methods to identify truth from
fallacy, denounced it as delusional. There is no scientific evidence
behind it at all. The general idea was, however, turned into very
many highly popular music hall tricks. In one popular version, the
performer identified which name was the name of a dead person
from a set of random names offered by the audience. One way to do
this is to have the first person write the name of a dead person on
their piece of paper, while the other three write the names of living
people on theirs, then do the trick as above. While psychometry
is nonsense, Jastrow, who denounced it, was a real scientist who
discovered and popularised lots of optical illusions as part of his
actual scientific work.
 Chapter 26

The Roman Maths Challenge

Roman Numerals

Conjuring
In this challenge, the audience must do some Roman numeral
maths, making a Roman numeral sum add up.

Computation
The way you choose to represent numbers affects how hard or easy
it is to do different tasks. Addition and subtraction are fairly easy
with Roman numerals but multiplication and division are harder.
When choosing a representation of data, it is important to think
about what you will need to do with the data.

The Trick
Explain that this is more a challenge than a magic trick. Ask a friend
if they know Roman numerals. Remind them that Roman numerals are
just a different way of writing numbers, though one that makes doing
some maths harder than ours. They used I to mean one, V for five,
X for 10, counting I, II, III, IV, V, VI, VII and so on. Say “Let’s do some
Roman maths. . . ” On a pad, slowly write out the start of a sum using

171
172 Conjuring with Computation

Roman numerals, saying each number after you have written it to check
they are the following:
II + II = IV
“Two . . . plus . . . two . . . equals . . . four”
II + IV =
“Two . . . plus . . . four . . . equals . . . ”
Next, write in the Roman numeral for 9 (IX) saying,
“This is an obviously wrong bit of maths using Roman numerals”:
II + IV = IX
Finally, lay the challenge: “The challenge is: can you make it work so
that the two sides are equal by adding just a single line?”
Give them a few minutes to try, and if they fail, then show them the
answer. (If they manage to do it, praise them for their brilliance!)

How It Works
The solution is to draw in an S before the IX to make the correct sum.
II + IV = SIX
Point out that you didn’t say they had to use a straight line! You also
didn’t say they had to stick to Roman numerals.

The Magical Algorithm

TO DO The Roman Maths Challenge:
1. Explain how Roman numerals represent numbers.
2. Write II + II = IV while saying “Two . . . plus . . . two . . . equals . . .
four”.
3. Write II + IV = while saying “Two . . . plus . . . four . . . equals . . . ”.
4. Finish the sum by writing IX to give II + IV = IX.
5. Say “This is an obviously wrong bit of maths using Roman
numerals”.
 The Roman Maths Challenge: Roman Numerals 173

6. Say “The challenge is: can you make it work so that the two sides
are equal by adding just a single line?”.
7. Give them a few minutes to work it out.
8. Write an “S” before the IX to make the sum work.
9. Say “I didn’t say you had to use a straight line!”.

Thinking Computationally
When we think of numbers, we tend to think of the digits we use to
write them, but that is just a representation: one particular way of
writing them. There are lots of different ways to write numbers. Early
civilisations just used tally marks: a very direct representation with one
mark for each thing you were counting. If you have 12 goats, then you
would write 12 lines.
IIIIIIIIIIII
This representation is called unary. It is fine for small numbers, but
it gets harder to keep track of what number it is for larger numbers.
The first improvement is to make the groups of five more obvious
by clustering them, such as making every fifth line into a bar. This
representation is called tallying.

IIII  II
IIII
That makes keeping track of how many you have easier as you can check
by counting the tallies in fives. Adding two numbers represented as tallies
is fairly easy too for small numbers as you just combine the two lists
of tallies, and then arrange any extra ones into clumps of five. It also
works well given a limitation of the way our brains work. Scatter some
number of coins on the table. If there are less than five or so, you can
probably see how many are there at a glance. If there are more though,
then you will start having to count to tell. Our brains are good at just
seeing small numbers, but can’t do it naturally for larger ones. You can
probably see how many there are in this sum:
II + III = II III
But, what number is the following? You have to count them all over
again! It’s just marks you are counting now instead of goats.
IIIIIIIIIIIIIIIIIIII
174 Conjuring with Computation

Roman numerals are a different way of representing numbers that

solves this problem by introducing new symbols to use in ever bigger
numbers. It is a different representation that makes quickly seeing what
number you have easy, even for really large numbers. The above number
is written as
XX
which is much easier to deal with. You can see how many Xs there are
at a glance.
However, doing arithmetic in Roman numerals is quite hard. Adding
isn’t so bad, but multiplication is really hard, by comparison. That’s why
we no longer use Roman numerals. Now we use a different representation
of numbers that in particular is convenient for writing arbitrarily large
numbers and also that makes doing all kinds of arithmetic much easier,
so is better all around. We use a representation that makes the job easier.
Computers use a different representation again, based on the one we
use, but that is a more convenient version for the way computers work.
We will look at these other representations in subsequent chapters.

Interlude: Nevil Maskelyne

Nevil Maskelyne (1863–1924) was the son of a famous magician,
John Nevil Maskelyne. He became a magician in his own right
making use of new communications technology in his mentalism
tricks. When Guglielmo Marconi’s wireless telegraph (i.e., radio)
seemed to be set to pull the plug on the cable telegraph, one of
the cable companies hired Maskelyne to spy on Marconi listening
in to all his broadcasts. Maskelyne did more than this, debunking
Marconi’s claims very publicly that radio was private and could not
be hacked. He set up his own broadcast station in London and used it
to ruin Marconi’s prestigious public demonstration of the technology
at the Royal Institution by sending his own message picked up by the
demonstration receivers before those of Marconi. His fake messages
poked fun at Marconi.
 Chapter 27

The Lottery Trick

Place-Value Number Representation

Conjuring
Everyone chooses a lucky lottery number. You then turn the room
into a lottery machine with everyone bouncing lottery balls around.
Three numbers are chosen at random and added, and then we see
if anyone has a lucky lottery ticket. The lucky winner is of course
you, the magician.

Computation
We use the Hindu–Arabic numeral system. The really amazing idea
here is that the value of digits changes depending on their position
in the number, so 2 can stand for two, twenty or two hundred
depending on its position. This makes large numbers more concise
but also makes operations like addition and multiplication easier.
It makes for simple algorithms.

175
176 Conjuring with Computation

The Trick
Before the trick, write the numbers 3, 4 and 8 each on identical pieces
of red paper, the numbers 1, 5 and 9 on yellow paper, and the numbers
2, 6 and 7 on blue paper. Secretly, write the number 1,665 on a piece of
paper made to look like a lottery ticket and put it in a sealed envelope.
Explain to the audience that you are going to play a lottery game, but
where the whole audience will be the lottery machine. Everyone should
write down their lucky 4-digit number on a piece of paper and place it
in an envelope. It will be their lottery ticket. Hold up your envelope and
say you will play too and that the envelope contains your lottery ticket.
Tape it to the wall so all can see it, with either masking tape or some
other way that will not damage the wall.
First, you need to randomly choose the order of the colours that will
determine the order that balls are removed from the lottery machine.
Have a volunteer decide the order of the three colours (red, yellow and
blue) at random to use. Write the colours in the order they decide on a
flip chart.
Explain that you are going to make the lottery balls out of screwed-
up coloured paper so they can bounce around the room without hurting
anyone. Show the audience the pieces of paper with the numbers. Explain
that they will be the lottery balls, noting that just like lottery balls they
are coloured and have numbers. You have used three colours and the
numbers one to nine.
Next, ask a volunteer to screw each piece of paper up into a ball and
toss them into the audience. The audience should then toss the balls of
paper around as if they are in a lottery machine.
Next, go to the order of colours chosen, and ask the volunteer to
pick out one ball of the first colour. Unscrunch it to find its number.
Write this under the colour. Do the same for the next colour, and then
the third. Explain you have the first 3-digit random number but that
you are going to do this three times and just add the numbers so that
all the balls are used. Repeat this process with the same order of colours
to get a second 3-digit number and finally a third. You now have three
3-digit numbers (see Figure 27.1).
 The Lottery Trick: Place-Value Number Representation 177

Figure 27.1. This shows what would be written on the flip chart if the colour order
chosen was: blue, then yellow, then red; and then balls are drawn randomly as follows:
Blue — 6, Yellow — 9, Red — 3, Blue — 2, Yellow — 5, Red — 8, Blue — 7,
Yellow — 1 and Red — 4. The numbers 693, 258 and 714 are added on a calculator
to give the answer and so the final lottery number, 1,665.

Give the volunteer a calculator and have them add up the three
numbers. Then, turning to the next sheet on the flip chart, write the
total and circle it. Announce that is the winning number and ask if
anyone has won. Finally, have the volunteer open the envelope to reveal
your lottery ticket and read out your number. You have indeed picked
the winning number, 1,665!

How It Works
The choice of order of colours appears to add randomness to the process,
but it actually makes no difference. Let us look at just one colour, red.
The three numbers there are 3, 4 and 8. Whichever position red is put
in, those three numbers will always be in that same column as that is
where the red numbers are written. That means that even though the
order they are chosen changes, they will always be added together (and
they add up to 15). The three numbers on the other coloured pieces of
paper will also always be added together, whatever order the balls are
pulled out. The yellow numbers, 1, 5 and 9 will be added also giving 15.
Similarly, the blue numbers 2, 6 and 7 will be added together and add
up to 15 too.
This means that whatever order the balls are chosen, the three
columns of numbers will add up to 15. You will always be adding 15, 150
and 1,500 as each represents a different decimal place so is increased
178 Conjuring with Computation

by 10 from the previous one. 1,500, 150 and 15 add up to 1,665, the
number you wrote on your lottery ticket.

The Magical Algorithm

TO PREPARE The Lottery Trick:
1. Write the numbers 3, 4 and 8 each on a red piece of A4 paper.
2. Write the numbers 1, 5 and 9 each on a yellow piece of A4 paper.
3. Write the numbers 2, 6 and 7 each on a blue piece of A4 paper.
4. Write the number 1,665 on a white piece of paper drawn to look
like a lottery ticket.
5. Seal the white piece of paper in an envelope.

TO DO The Lottery Trick:

1. Have everyone in the audience write down a lucky 4-digit number
of their choice and seal it in an envelope.
2. Hold up your envelope noting it contains your lottery number
and tape it to the wall.
3. Ask a volunteer to choose the order of colours to pull out balls
from the lottery machine from a choice of red, yellow and blue.
4. Write the colours down in the order they choose on a flip chart.
5. Have the volunteer screw up the 9 pieces of coloured paper and
throw them into the audience one at a time.
6. The audience toss the balls around as though they are balls in a
lottery machine.
7. DO THE FOLLOWING 9 TIMES.
a. Announce the next colour in turn as written on the board.
b. Ask the volunteer to choose a ball of paper of that colour
from those in the room.
c. Have the volunteer unscrew the ball and announce the
number.
d. Write the number on the board in a column under the colour.
 The Lottery Trick: Place-Value Number Representation 179

8. Ask the volunteer to add up the three 3-digit numbers under the
colours using a calculator.
9. Turn to the next sheet on the flip chart and have them write the
total there and circle it.
10. Announce that that number is the winning number.
11. Ask anyone with a winning ticket to claim the prize.
12. Have the volunteer open your envelope to reveal that you have
a winning ticket.

Thinking Computationally
This trick relies on the fact that we use the Hindu–Arabic number
system. It was invented by Indian mathematicians in the first few
centuries BC but brought to the West by Arab scholars and traders.
The really amazing idea that matters most about Hindu–Arabic numbers
is that digits have a position in a number. Unlike in Roman numerals
(where for example X means 10 wherever it appears), their value changes
depending on their position. For example, in the Hindu–Arabic system,
2 can stand for 2, 20, or 200, and so on, depending on its position in
the number. This is a vital part of the trick that makes it work. The
numbers are placed in a position labelled by their colour so that they all
come back together to be added.
This representation makes large numbers very concise, so taking up
little memory in computer terms. It also makes manipulating numbers
by humans relatively easy: it makes for relatively simple algorithms such
as the ones we learn in primary school. Counting is easy, for example, as
are doing things like addition and multiplication. Multiplication is much
easier than with Roman numerals. The choice of representation makes
tasks easier or harder.
This is the key idea about choosing a good representation for data,
whether numbers or other kinds of data: you want to make it easy to do
the key operations. With computers, that usually means you want there
to be fast algorithms, though sometimes taking up as little memory as
possible matters more, and sometimes you just want it to be easy to get
the algorithm right.
180 Conjuring with Computation

PAUSE: Charles Babbage

Charles Babbage (1791–1871) designed the first computer, the
Analytical Engine, though never actually managed to build one. It
was a mechanical computer with punch cards based on the Jacquard
Loom to provide the program. It would have been steam-driven, and
the data was stored in a series of cog-like wheels. Unlike modern
computers, its data was to be stored as decimal numbers based
on the positions of the wheels, so using the Hindu–Arabic number
representation. Such wheels were used in his earlier and simpler
Difference Engine, a prototype of which was built. He was not a fan
of all machines: he hated barrel organs (or at least the people playing
them in the street) so much he tried to get an act of parliament to
outlaw them. A mechanism based on a barrel organ was included in
his design of the Analytical Engine.
 Chapter 28

Cards on Your Mind

How Computers Represent Numbers

Conjuring
Four volunteers jointly pick a number and jointly concentrate on
it, helped by your set of magic cards. Despite them keeping it
secret, you are able to read their collective mind and announce
their secret number.

Computation
Digital computers do not use the decimal system that we use but
instead use a different place-value system called binary. Binary is
also the underlying representation that makes certain tricks work.

The Trick
Ask four volunteers to help in an experiment in mind reading. Give them
a list of 15 numbers and ask them to jointly pick one without telling you
what it is. Then give each a card which contains a jumble of numbers
and letters. They stand in a line facing you, cards hidden from your sight.
They each look to see if the jointly chosen number is on their card. If it
is, then they must close their eyes and think hard only of that number.

181
182 Conjuring with Computation

Those holding a card without the number choose one of the letters on
their card instead and stare at it, thinking only of that letter. That way
they won’t confuse you, you say. Once they are all concentrating, you
close your eyes, think hard, then pick up the thoughts about the secret
number and announce what their secret number was!

How It Works
You will need the set of 5 cards shown in Figures 28.1 and 28.2. You can
download them from https://conjuringwithcomputation.wordpress.com,
print them out, and stick them onto card or write out your own with
the same numbers and letters on.
First, you give four volunteers the card with the numbers 1 to 15
to look at (Figure 28.1). They jointly choose one and keep it secret.
Next, you give them the 4 special cards with numbers and letters on,
one each, in the order as shown in Figure 28.2.

Figure 28.1. The card with the numbers 1 to 15 on, for the “Cards on Your Mind”
trick.
 Cards on Your Mind: How Computers Represent Numbers 183

Figure 28.2. The 4 cards to give out for the “Cards on Your Mind” trick.

These cards are special in that the numbers on each are carefully
selected. Secretly think of them as card 8, card 4, card 2 and card 1.
These are the numbers in the top left corner of the card. Stand the
volunteers in a line and give out the cards in that order, so the person
on the right facing you has card 1 and the person on the left has card 8.
Now, get those with their jointly chosen number on their card to
close their eyes. You add up the card numbers of the people closing
their eyes, and that is their secret number. For example, if the people
holding card 1 and card 8 have closed their eyes and the other two
volunteers are staring at their card, eyes open, then add 1 and 8 and
announce they are thinking of number 9. If instead the people holding
card 2, card 4 and card 8 have their eyes closed, then the secret number
is 2 + 4 + 8 so 14. It is that simple. You don’t need to be able to read
minds, just add up powers of 2.

The Magical Algorithm

TO DO Cards on Your Mind:
1. Give four volunteers the card containing numbers 1 to 15
(Figure 28.1).
2. Ask them to jointly pick one number between them without telling
you what it is.
3. Have them stand in a line facing you.
184 Conjuring with Computation

4. Give each a card which contains a jumble of numbers and letters in

the order of the cards based on the number in the top left corner,
card 1 on the right, card 8 on the left as shown in Figure 28.2.
5. Tell them to keep the cards hidden from your sight and each look
to see if the jointly chosen number is on their card.
6. IF the chosen number is on their card
THEN they should shut their eyes and think only of that number.
OTHERWISE they should choose a letter on their card and think
only of that letter so as not to confuse you.
7. You read off the number by adding the card numbers for the eyes-
closed people to give the total.
8. Announce that number as the secret number.

Thinking Computationally
The secret of this mind-reading trick relies on the binary representation
of numbers. Figure 28.3 gives numbers from 1 to 15 with their binary
version. It is a way of writing numbers using sequences of 1s and 0s. It
is a positional number system that works just like decimal, but where
there are only two symbols: 0 and 1 instead of 10, and the positions
increase in powers of 2 (1, 2, 4, 8, . . .) instead of powers of 10 (1, 10,
100, 1,000, . . .).
To work out the value represented by the number in both decimal
and binary, you just multiply the number in a column by the column
heading and then add the answers. In decimal, the column headings are
powers of 10 (as there are 10 symbols). In binary, the column headings
are powers of 2 (as there are 2 symbols).
That means the 1 in the binary number 10 means 2, the 1 in the
binary number 100 means 4, and the 1 in the binary number 1,000 means
8. So just as the number 1,204 in decimal represents 1 × 1,000 + 2 ×
100 + 0 × 10 + 4 × 1, the binary number 1,101 represents 1 × 8 +
1 × 4 + 0 × 2 + 1 × 1 (so means 13 as 8 + 4 + 1 = 13).
You can use any two different things to represent those 1s and 0s
as they are just arbitrary symbols. In the trick, eyes open is used for 0
and eyes closed is used for 1. The secret number is then being spelled
 Cards on Your Mind: How Computers Represent Numbers 185

8421
1 0001 = 0+0+0+1
2 0010 = 0+0+2+0
3 0011 = 0+0+2+1
4 0100 = 0+4+0+0
5 0101 = 0+4+0+1
6 0110 = 0+4+2+0
7 0111 = 0+4+2+1
8 1000 = 8+0+0+0
9 1001 = 8+0+0+1
10 1010 = 8+0+2+0
11 1011 = 8+0+2+1
12 1100 = 8+4+0+0
13 1101 = 8+4+0+1
14 1110 = 8+4+2+0
15 1111 = 8+4+2+1
Figure 28.3. The numbers 1 to 15 with their binary representation and as an addition
of powers of 2 corresponding to the positions of 1s.

out in binary by the four people’s eyes. You can just read the binary to
know what the chosen number was.
This works because the numbers on the cards are just those with a
1 in the corresponding position of that number when written as binary.
Card 1, for example has all the odd numbers written on it as in binary
all odd numbers have a 1 in the first position (the 1s column). See
Figure 28.3 and notice how the first column has a 1 for numbers 1, 3,
5, . . .
Similarly, card 8, the leftmost card, has the numbers from 8 to 15
written on it as those numbers have a 1 in the fourth position (the 8s
column) in the binary.
By telling people to close their eyes if the number is on the card you
are secretly saying close your eyes if there is a 1 in the binary at your
position. The volunteers are representing their chosen binary number.
Adding up powers of 2, for eyes shut or not, is then just the way you
work out what number that binary number actually represents.
You don’t actually have to remember the table of Figure 28.3 to
work out what decimal number the binary number stands for. You can
just do the calculation instead to work out what number your volunteers
are thinking of. We put the numbers 8, 4, 2 and 1 above the columns,
and you should think of the person in each position corresponding to
186 Conjuring with Computation

that number. You then just add up the column numbers to get the secret
number. If there is a 1 (eyes shut) in the binary in a column, then that
column number is added to the total.
For computers, binary is a better representation of numbers than
decimal. In the trick, the 1s and 0s are represented by eyes shut or eyes
open. In a computer, they are just differing electrical signals: high and
low voltages on wires or transistors that are on or off, for example. These
binary signals are easily stored and easily transferred from place to place
in the computer, meaning the binary numbers can be moved to where
needed to do a calculation. There are also very fast algorithms for doing
arithmetic, and it is easy to create specialist hardware for computers
to use to do the maths. This all makes binary an ideal representation
for numbers in a computer. Therefore, all data are ultimately stored in
computers as binary 1s and 0s rather than as decimal numbers.

PAUSE: Gottfried Leibniz

German philosopher and mathematician Gottfried Leibniz (1646–
1716) (amongst others) studied the use of binary numbers in
the 17th century. In part, he was fascinated by the symbols
in a 9th-century Chinese text (a version of the I Ching) which
shows that binary existed long before its western uses. A form
of binary arithmetic was also used even earlier in Ancient Egypt.
Leibniz worked out the mathematics behind binary. He was also
an important logician setting out the basics of formal logic which
is the basis of logical reasoning. He worked on machines to do
calculations and even imagined a machine controlled by punch cards
with binary numbers represented by marbles long before Charles
Babbage envisioned his punch card-controlled machines.
 Chapter 29

On the Power of Numbers

Representing Other Things by Numbers

Conjuring
A postcard is chosen from a selection while your guest magician is
out of the room, but despite that they can quickly say which card
was touched.

Computation
All sorts of things can be represented by numbers and once they
are then we can bring them into the digital world, as those numbers
in turn can be represented by binary.

The Trick
Explain that your guest magician has extremely heightened powers of
observation and can tell when an object such as one of the set of nine
postcards on the table has been touched by someone. Have them leave
the room, then wearing gloves, take the postcards and shuffle them. Lay
them out on the table in a rough three-by-three grid and ask a volunteer
to pick one by touching it. They should then remember their card.

187
188 Conjuring with Computation

Collect the cards back up, shuffle them, and then lay them back
down on the table, still wearing gloves. Call in your partner magician.
Immediately, just by looking at the cards, they are able to say which
card was touched.

How It Works
In advance, you agree a number from 1 to 9 for each card. Both memorise
these numbers for the cards. To make this easier, you could choose cards
that have different numbers of objects: perhaps the one you label 1 is
a picture of a single lighthouse, whereas the one labelled 3 has three
seagulls flying in the background, and so on.
As your partner enters the room, facing them, you casually put your
hands together, touching that many fingers together as you do so. If you
do this as your accomplice enters the room, announcing their arrival, the
audience are likely looking at the door not at you to make it even less
likely they will see or be suspicious about the movement of your hands.
A variation, rather than labelling the cards (should you find that hard
to remember), would be to communicate the position of the chosen card
as a pair of numbers giving an x and y coordinate with the fingers you
touch together on each hand. So if the card was in the bottom left, you
would touch one finger on each hand. If in the top left, touch one finger
to three fingers.

The Magical Algorithm

TO DO the Preparation of On the power of numbers:
1. Choose 9 postcards with different pictures.
2. Agree with your partner a number for each picture.

TO DO On the power of numbers:

1. Explain your partner magician has heightened sensory powers and
can tell when an object has been touched.
2. Your partner magician leaves the room.
3. Put on gloves and lay the cards out on the table in a rough 3 × 3
grid.
 On the Power of Numbers: Representing Other Things by Numbers 189

4. Ask a volunteer to choose one by touching it and then everyone

remembers their chosen card.
5. Take up the cards and shuffle them.
6. Deal them back into a rough 3 × 3 grid.
7. Call your partner magician back in.
8. Touch together the number of fingers as the number of the chosen
card so that your partner can see it.
9. Your partner picks up the card corresponding to that number
saying it was the one that had been touched by someone.

Thinking Computationally
We have ways to represent numbers in binary, but numbers themselves
can be used to represent other things: cards and the images on them
in a simple way as part of this trick. The same can be done for music,
images, money . . . just about anything. That idea is the foundation for
the digital world.
Images do not have to be represented just by a count of the objects
in them. The actual image can be turned into numbers that describe
the image in detail. There are several ways to do this. One way, called
vector graphics, involves describing each line or shape in the image
by its size and position. Another, called raster graphics or bitmap
graphics, involves breaking the image into small squares and giving
numbers to each square.
To do this, first, we need to represent colours as numbers. One
simple way is just to give each colour in the image a single number: red
is 1, orange is 2, yellow is 3, and so on. Next divide the image into lots
of small squares and note which colour is the main one in that square.
Represent that square (or pixel) by the number for that colour. Now the
list of numbers for all those pixels is a representation of the picture. Each
number in turn can be represented by binary so stored on a computer or
sent over a network. By sending the list of numbers as binary over the
Internet or phone network to someone else, they can reconstruct a copy
of the image.
If an image or piece of music or a film or a book can be represented
by numbers like this, then it can be converted into a digital object.
190 Conjuring with Computation

Then, it can be copied, shared, streamed . . . just by copying, sharing,

streaming . . . those numbers.

Interlude: Luca Pacioli

The oldest known book that describes card tricks, De viribus
quantitatis or On the Power of Numbers, included a variation of this
trick of the same name. It was a book of mathematical diversions
written by Luca Pacioli (c.1447–1517) around 1500. It suggested
that a conjurer could train an accomplice to “guess” which card a
person had touched, even though they were out of the room at the
time, by agreeing in advance numbers linked to the cards. Just like
many modern magicians, Pacioli was keen to show that often effects
that were classed as magic were really just conjuring tricks. Pacioli
was a close friend of the artist and engineer, Leonardo da Vinci,
and taught him mathematics as an adult. Leonardo had realised
maths would help him gain both a better artistic and scientific
understanding of the world. Leonardo’s only published illustrations
(of three-dimensional shapes) were illustrations for another book by
Pacioli.
 Chapter 30

The Out-of-Body Experience

Error-Correcting Codes

Conjuring
Someone flips over a card in a grid of cards. Blindfolded, you can
leave your body, and hovering from the ceiling watch what happens
and so can immediately say which card was changed once you
return and the blindfold is removed.

Computation
Computers use special representations to send data over networks.
Error-correcting codes are ways to represent, and so store and
transmit, data so that if some are corrupted, a computer can not
only tell it has been changed but also correct the changes. It can
do this without knowing in advance what the original data was.

The Trick
For this trick, you need a partner magician. Explain to the audience that
you are able to float out of your body and see things you would otherwise
not be able to see. You go to the back of the room where a volunteer
blindfolds you. Have them check the blindfold first to ensure there is

191
192 Conjuring with Computation

no trickery. They stay with you throughout the whole trick to both
“guard your body” and make sure you remain blindfolded throughout.
You explain that your spirit is about to leave your body and as you do
so your body slumps.
Once you are blindfolded and spirit departed, your partner magician
shuffles a pack of cards. A volunteer then deals them out in a 4 × 4
square. They place some cards face up and some face down at random.
Your partner suggests that to make it harder they will make the square
even bigger and turn it into a 5 × 5 square which they do by adding
more cards at random.
You are blindfolded and lean against the wall at the back of the
room with your back to the proceedings through all of this. Your spirit,
of course, left your body and flies up to the ceiling so you can watch
from above. It has a target to watch in the square of cards. Next a
new volunteer chooses any card from the grid and flips it over. No one
speaks. You are still blindfolded. You can only know which one was
flipped if your spirit really is floating above, watching. You are told by
your partner to return to your body, which you do. As you return, you
bang against the wall as you re-enter too quickly. As you were upside
down on the ceiling, you will be a bit dizzy when you return so you may
wobble about a bit. Return to the front and look at the square of cards.
You have trouble working out which way up the square was if you were
upside down, so struggle to work it out. You turn your head to one side
before finally pointing to the one that was flipped over!

How It Works
The secret is in that extra row and column of cards your partner adds
apparently at random (see Figure 30.1). It isn’t in fact random. It also
isn’t making things harder, but easier. Your partner counts the number
of face-down cards in each row. If that number is odd so that either 1
or 3 of the cards is face down, they put the new card face down. With
that added card, there is now an even number of cards (2 or 4) face
down in the row. If the number of face-down cards in the row is already
even (0, 2 or 4), then instead they put the new card face up. This also
leaves the number even.
 The Out-of-Body Experience: Error-Correcting Codes 193

(a)

(b)

(c)

Figure 30.1. (a) An initial layout of cards, some face up (white), some face down
(grey). (b) The extra cards added to the rows and columns make each have an even
number of face-down cards. (c) When one is turned over, it can be identified by
finding the row and column with an odd number of face-down (grey) cards.

Once they have done this for all rows, every row has an even number
of face-down cards.
Next they add cards to the bottom of the columns, and do this in
exactly the same way, ensuring all columns end up with an even number
194 Conjuring with Computation

of face-down cards, including the last card in the corner. That ensures
that the extra column of cards added to each row also ends up with an
even number of face-down cards.
Working out which card has been turned over is now easy as long as
only one card is changed. The row that the turned card is in will now have
an odd number of face-down cards. Similarly, the column containing the
turned-over card will also have an odd number of face-down cards. This
is because in each case one has been turned over either turning a face-up
card to face-down or vice versa. Turning one card either way means the
even pattern is broken but only in one row and one column.
Where the row and column with odd numbers of cards meet is the
card that was turned.
The trick works just as well, of course, however big the square of
cards is. It does not have to be a 5 × 5 square.

The Magical Algorithm

TO DO The Out-of-Body Experience:
1. Explain to the audience that you are able to float out of your
body and your spirit self then sees things.
2. You are blindfolded at the back of the room and “leave your
body.”
3. Your partner shuffles a pack of cards.
4. A volunteer deals them out in a 4 × 4 grid placing cards face up
or face down at random.
5. Your partner says they will make the square bigger so it is harder.
6. FOR each row of cards in the grid
a. IF there are an odd number of face-down cards in the row,
THEN your partner adds a face-down card at the end of the
row.
OTHERWISE your partner adds an extra face-up card at the
end of the row.
7. FOR each column of cards in the grid
a. IF there are an odd number of face-down cards in the column,
THEN your partner adds a face-down card at the end of the
column.
OTHERWISE your partner adds an extra face-up card at the
end of the column.
 The Out-of-Body Experience: Error-Correcting Codes 195

8. A second volunteer picks a card and flips it over.

9. Your partner tells you to return to your body and remove your
blindfold.
10. You remove the blindfold and come to the front to look at the
grid.
11. Point out that you did not see it before the flip, so have no way
of knowing what card was changed.
12. Identify the row containing an odd number of face-down cards.
13. Identify the column containing an odd number of face-down
cards.
14. Pick up the card that is in both the identified row and identified
column and announce that it is the card flipped.

Thinking Computationally
We are using a clever, more advanced kind of data representation here,
called an error-checking code. They allow us to spot when mistakes
have been made, and for some codes like this one, allow the mistakes
to be corrected too. If they can correct the error, as here, then they are
called error-correcting codes.
Everything in a computer is ultimately stored as binary bits:
sequences of 1s and 0s. That is also how data is transmitted. If you
stream music, the music is represented as a long stream of numbers,
but those numbers are each a series of 1s and 0s. Those bits can easily
be corrupted in transit: whether zapped by cosmic rays or just corrupted
by other electrical interference along the way. If bits are flipped by all
this noise, then the music you receive will be corrupted. For some kinds
of data, one change could make the message meaningless. We need a
way to deal with the problem. Rather than protecting the message as
it travels over the network, error-correcting codes allow us to just fix
problems when the data arrives.
Instead of thinking about face-up and face-down cards in the trick,
now think about binary 1 and 0. The square of cards could just as easily
represent 16 bits (16 1s or 0s) of computer data. Take 16 bits and
organise them in a square and then add the extra bits to make each row
and column an even number of 1s. Computer scientists call these extra
196 Conjuring with Computation

(a)

(b)

(c)

(d)

(e)

(f)

(g)

Figure 30.2. (a) The message to send. (b) Those bits laid out in a square. (c) Extra
parity bits are added to each row or column. (d) The longer message is sent. (e) A
version of that message is received with the 12th bit flipped. (f) The received bits are
laid out in a square, and the corrupted bit is identified. (g) The message is ultimately
received after the correction.
 The Out-of-Body Experience: Error-Correcting Codes 197

bits parity bits. Now you have a slightly longer message of 25 bits that
you can send instead of the original 16.
At the other end, the receiver of those 25 bits can put them into a
square and check all the parity bits are correct. If any row or column has
odd parity, then a bit was corrupted. If there is one row and one column
wrong, then the flipped bit can be determined just as in the trick, and
the mistake can just be flipped back (see Figure 30.2).
If more than one bit was flipped, the receiver at least sometimes also
knows the data was corrupted and can ask for it to be sent again. If you
have more data to send than 16 bits, then you can just split it up into
blocks of bits and add the parity bits to each. The receiver checks each
in turn. Blocks can also of course be bigger than 16 bits as long as it is
a square number. The best size of blocks will depend on how likely it is
that bits will be corrupted.
This error-correcting code is a very simple one. There are much more
powerful representations that can be used to both detect and correct
multiple corrupted bits in data.
Without a variation of the parity trick to help, all digital data
transmitted across the Internet, from text messages, websites to emails,
or streamed music and films, would be full of errors. Given how integral
digital data is now to our lives, it is a pretty important trick.

Interlude: Tim Bell

The “Out-of-Body Experience” trick is a variation of one invented by
New Zealand computer scientist, Tim Bell. He pioneered the idea of
using conjuring tricks to illustrate complex computer science ideas.
He used it, along with other tricks, games and activities, to inspire
primary school children about computer science as part of his “CS
Unplugged” program. It is the earliest example we are aware of, of
a trick being specifically invented to teach computer science, here
error-correcting codes.
This page intentionally left blank
 Part 8

Human–Computer Interaction

We now return to the presentation side of tricks and the importance

of understanding people (i.e., cognitive psychology) when developing
computer systems. In computing terms, this presentation side is called
Human–Computer Interaction or Interaction Design. It is about ensuring
programs are usable (usability) and provide a good user experience.
This page intentionally left blank
 Chapter 31

The Four Aces

Usability and Attention

Conjuring
The audience try to keep track of the Aces in a game of cards.
The person who clearly held all the Aces turns out to have nothing.
Instead, you, the magician, have them all.

Computation
Programmers must make their programs easy to use so that
people do not make mistakes using them. Programmers must take
our limitations into account. One such limitation is our focus of
attention. Programs should actively control our attention to help
us both avoid mistakes and recover from mistakes when we do
make them.

The Trick
Get a volunteer to the front, and have everyone else gather around the
table. Explain that this is a trick about why you should never gamble, and
certainly not with magicians or computer scientists. Place the four Aces

201
202 Conjuring with Computation

Figure 31.1. The initial position of the four Aces.

from a pack of cards face up onto the table to start four piles. Add three
random normal cards face down on top of each Ace showing them to
the audience as you do so (see Figure 31.1).
Explain that you have dealt out four hands for the purposes of
demonstration. Everyone should imagine that a game had just been
played and this was the way the cards fell and that you just happened to
notice where the Aces had ended up. In this game, the person with the
most Aces wins. That means that anyone who can keep track of where
the Aces go will have an advantage and know when to stay in the game
and when to drop out, when to gamble big and when not to gamble at
all. Everyone should therefore watch where the Aces are at all times.
You turn each Ace face down leaving it at the bottom of its pile
and pick up the four piles in order. Point out that if the dealer does
not shuffle the cards, then you know exactly where the Aces are and
so where they will go on the next deal. They are every fourth card in
the pack, so when four new hands are dealt out, all four Aces will land
together in the fourth pile. That person will have a winning hand. Start
to deal out the cards to illustrate it happening. Count: “1, 2, 3 and an
Ace” as you deal the cards face down on the table. On dealing the Ace,
stop and, pointing to the Ace with the next card from the deck, ask the
volunteer to show the audience it really is an Ace, so all can see that you
are definitely right and that person does get the Aces. Have them then
leave the Ace face up on the table so everyone can see that that is where
 The Four Aces: Usability and Attention 203

the Aces go. Continue to deal the other cards out, counting “1, 2, 3 and
an Ace” as you do to emphasise that the Aces end up together in the
fourth pile.
Point out that the fourth person (the volunteer) has all the Aces so
will win. If this was a gambling game, they would put lots of money on
the table. They will keep raising the stakes as long as anyone is willing
to match them as they can’t lose.
Two hands are then eliminated, perhaps because like us they know
where the Aces are, or perhaps they just cannot afford to lose that
amount of money. The third hand is yours, so you really ought to drop
out too if you had any sense.
In this game, any player still in the game can now swap a card with
another player if they wish. The person with all the Aces will pass as it
would make their hand worse. You, the magician, do swap a card. Take
the face-up Ace from the bottom of the fourth pile and swap it with the
bottom card of your pile, turning the other card over to show there is
no trickery. Note that you have gained an Ace, but the other hand still
wins as it has three Aces left, pointing to the three face-down Aces as
you say this. Tell everyone that you really can’t win so ought to fold.
Instead, you are going to steal those Aces with everyone watching.
Tell them to keep their eyes on the fourth pile. You will make them
all look away at just the moment when you steal the Aces. On a count
of three, clap your hands in front of the volunteer’s face.
Ask if anyone saw you steal the Aces!
Point out that the volunteer definitely had three Aces and a winning
hand, but now . . . turn the volunteer’s pile over . . . The Aces are gone.
All that money gambled has been lost! So where did the Aces go? Turn
over your pile to reveal that you now have all the Aces. You really did
steal them without anyone seeing.
That is why you should never gamble, and certainly not with
magicians or computer scientists!

How It Works
The secret behind this trick is that you use misdirection, making the whole
audience miss something important even though it is there to be seen.
204 Conjuring with Computation

When the volunteer shows the audience the Ace, they are being
misdirected. You pointed to that card to take their attention to it, and
the volunteer helps keep it there by turning it over and showing that it
is an Ace.
So what do you do that they all miss? You use the card that you
are holding to point at the Ace. As you stop pointing, you pull that card
back to where your other hand with the pack is. You slip that card to
the bottom of the pack and slide off the top card in its place. By the
time the audience’s attention is back to you, unbeknown to them, you
are holding a different card. More importantly, as the previous top card
is now at the bottom, the next Ace is now only three cards away from
the top, not four. That means as you continue to deal the cards, the
next Ace will fall on the third pile, not the fourth as everyone expects.
The third Ace is still four cards away, so it will also fall on this third pile
as will the final Ace. Only the Ace shown to the audience is actually in
the fourth pile. The others are placed in the third from the start, but
the audience don’t know that.
When you later swap the bottom card from the third pile with that of
the fourth, you are moving that lone Ace to join the others. Everything
from this point on is just performance to disguise what has already
happened.
The trick can be made more powerful by combining it with a false
choice (Chapter 21). Rather than choosing the two piles to drop out
yourself, let the volunteer apparently do so. After making clear the fourth
pile with the Aces is theirs, ask them to point to one of the other three.
If they point to the third, then you say that that hand will be yours and
the other two piles drop out. If they don’t, then you say the pile they
pointed to drops out. Then, using time misdirection, you talk about other
things for a while (like how everyone should drop out if they’ve tracked
the Aces not just the person that has). This is to help people forget
exactly what happened. Then, have them point to one of the remaining
two piles. Again whichever pile they point to, you ensure the third pile
stays and the other pile is removed. This makes it seem far more magical
that the Aces end up in that pile, making it seem less likely that the
misdirection happened when it did.
 The Four Aces: Usability and Attention 205

The Magical Algorithm

TO PREPARE The Four Aces:
1. Pull out the four Aces and 12 other random cards from a normal
pack of cards.

TO DO The Four Aces:

1. Place the 4 Aces face up on the table.
2. Place 3 other random cards face down on top of each.
3. Ask a volunteer to come to the front and stand by the fourth
pile.
4. Explain to the audience that the person with the most Aces wins
this game.
5. Collect up the four piles of cards, one pile at a time, making a
new pack, without shuffling.
6. Note the Aces are every fourth card in this new combined pile.
7. Explain that if you deal the cards now, all the Aces will land
together in the fourth pile and that you will now deal out cards
to demonstrate this.
8. Deal out the first four cards into four piles saying “1, 2, 3 and
an Ace”.
9. Pointing to the Ace with the next card, ask the volunteer to turn
it over to show everyone it is the Ace.
10. As the volunteer shows the Ace, place the card you were pointing
with on the bottom of the pile in your hand.
11. Pull the top card from the pack.
12. Ask the volunteer to leave the Ace face up so everyone can now
see that that is where the Aces will land.
13. Carry on dealing out the cards from the first pile while saying
“1, 2, 3 and an Ace. 1, 2, 3 and an Ace. 1, 2, 3 and an Ace”.
14. Point out that the volunteer has all the Aces so will win.
15. DO the False Choice WITH pile 3 chosen (allowing the volunteer
to appear to have a free choice over which of the first three piles
stay, but forcing it to be the third).
16. Say the remaining (third) pile is your hand and you should drop
out but instead are going to steal the Aces.
206 Conjuring with Computation

17. Introduce the rule that anyone left at this point can swap a card
with their opponent. You will, though the volunteer with all the
Aces won’t as it can only make their hand worse.
18. Turn over the bottom card of the third pile and swap it with the
face-up Ace that is in the fourth pile.
19. Note that the volunteer still has more Aces so will still win but
should watch those Aces closely as you are now going to make
the volunteer blink and steal the cards when they do.
20. Count to three, then clap loudly in front of the volunteer.
21. Ask if anyone saw you steal the Aces.
22. Turn over the fourth pile to show the Aces are gone.
23. Turn over the third pile to show you now have all the Aces.

Thinking Computationally (with Psychology

on the Side)
Just like all the other tricks, this one involves a series of steps that
must be followed. It has an underlying algorithm. However, it relies on
misdirection and some simple sleight of hand too. The audience do not
see everything. Presentation matters with all the tricks if they are to feel
magical. Here it is key.
The misdirection shows that we have a single focus of attention, and
when we focus our attention on an area, we do not see things nearby.
Pointing is a very strong social signal, and everyone is curious to check
the card is the Ace, so everyone looks at the card. For many in the
audience, the magician’s hands were not just there to be seen but in
their field of view, at least peripherally, at the time. They were looking
in the right general direction but still didn’t see the card being swapped.
Their attention was both moved to the card and also narrowly focussed
on it. When our attention is drawn to a point like this, we miss other
things even close by.
People make mistakes all the time, and it is not always because they
are stupid or negligent. We have limited cognitive resources like one focus
of attention. Magic tricks like this show that it is possible to engineer
a system so that everyone makes the same attentional mistake at the
same time. Magicians engineer the system so we look in the wrong place
 The Four Aces: Usability and Attention 207

and miss something important. Good user interface designers can use
the same understanding of cognitive science to make people look in the
right place at the right time. With good design, that takes account of
our limitations in the same way that a magician exploits them, designers
can help people avoid making mistakes or recover when they do.
Programming is not just about writing programs that work. It is also
about making them easy to use. Computer scientists call designing to
make things easier to use, usability. Many programmers write programs
with themselves in mind rather than the actual users who then find
them much harder to use than necessary. After all, the programmer
understands deeply how it works. The users do not. Many programs,
and features of programs, are so hard to use that they are never used.
Making a program easy to use is all about understanding the way our
brains work and what their strengths and limitations are.
Good programmers will want to know if people make mistakes with
their software and want to fix the problems. When people make mistakes
using their software, a good programmer will always ask, “How can I
improve it so that no one can possibly make that mistake again. If I
can’t, then how can I design it so they will always realise they made a
mistake and know immediately how to fix the problem.”
Some usability problems and the resulting mistakes are just mildly
irritating, like closing a program when you meant to minimise it, so
having to open it again. Others can lead to a lot of wasted time, like
forgetting to save the homework you’ve spent hours doing. Sometimes
the consequences of a mistake are impossible to undo: if you set your
alarm clock but didn’t realise the volume was muted, so miss an
interview, for example. In other situations, the consequences can be
about life or death. Hospitals are now full of computer-based gadgets
helping cure people or keeping them alive. Someone wrote the program
for every gadget. Someone did the interaction design (designed the way
users use the program), and so made it either easy to use, or conversely,
easy to make mistakes with, or most likely they made it somewhere in
between. If a nurse accidentally types in a dose that is 10 times too
much into an infusion pump, a machine pumping life-saving drugs into
a patient, then the mistake could kill them. High doses of otherwise
wonderful drugs are often toxic. Avoiding medical error matters.
208 Conjuring with Computation

In all the above cases, different designs can prevent the problems
or at least make them far less likely to happen and easier to spot and
correct. Word processors can save documents every few keystrokes rather
than leaving you to do it. That alarm clock could detect that the volume
is muted so know the set alarm will have no effect and ask you to fix the
problem before you finish setting it. The infusion pump could know the
safe range of doses of each drug and alert the person inputting it that
something is wrong.
Thinking about why a nurse might enter the wrong dose without
realising leads us back to our trick. There are many possible reasons,
but one issue is whether their attention was in the right place. If they are
typing in digits on a calculator-style digit keypad, then their attention
has to be on their fingers to see what they are doing, not on the screen
to see their mistake. When finished, their eye is likely drawn to the
start button. Perhaps we can pull the magicians’ trick and draw their
attention to the screen itself by making it flash so the nurse can check.
Even better, we can use a different way of entering numbers that means
the nurse’s attention stays on the screen as the numbers are entered.
One way to do that is to enter numbers using a cursor-based system.
For example, you might have two buttons that move a cursor left and
right from one digit to the next, and two more to change the value up
or down. Once the cursor is in the right place, a person’s attention is
then no longer on the buttons but on the screen watching the numbers
change, where it needs to be.
Controlling people’s attention is as important for a programmer as
it is for a magician.
Computers solve problems for people. If people are to use the
programs developed, then those programs must be designed to be usable.
They must be designed to be easy to use, despite the limitations of our
brains. Programmers must make sure mistakes are both hard to make
and easy to recover from.
Magicians design magic systems so that the whole audience make
the same mistake at the same time. Computer scientists must design
computer systems so that no one does if their systems are to be usable
and safe. They can use the same tricks, the same cognitive psychology,
to do so. Rather than directing a person’s attention away from things
 The Four Aces: Usability and Attention 209

Figure 31.2. Magic tricks and programs are composed of the same basic compo-
nents: an underlying algorithm or method and the way it is presented or interaction
design.

that matter, designers of software need to use the same understanding of

cognitive psychology to direct their attention to the things that matter
when they matter. Those tricks rely on an understanding of people as
well as technology. Both magic tricks and programs have the same two
equivalent important parts as shown in Figure 31.2: secret method and
presentation . . . algorithm and interaction design.

Interlude: Gustav Kuhn

Gustav Kuhn is a psychologist at Goldsmiths, University of London.
He investigates the science of magic and has, for example, used
eye-tracking technology, which works out where people are looking,
to investigate where the audience focus their attention when they
watch a magic trick. He found that misdirection is fascinatingly far
more complex than just making people look in the wrong place. Our
eyes can be looking directly at something, but we can still not be
aware of it at all.
This page intentionally left blank
 Chapter 32

Supernatural Suggestion
Visual Salience

Conjuring
You show five playing cards on the screen and ask everyone to
choose the one their eyes settle on. You then show the cards with
one removed: the one you think most people will have chosen.
When you ask, it turns out EVERYONE chose the removed card.

Computation
What you see depends on where you focus your attention, and
where you focus your attention depends partly on how salient the
things in the scene are. If you want important things to be seen
on a screen, then you have to make them visually salient.

The Trick
You explain that you are going to attempt an experiment in the power
of suggestion. Whilst any magician can make one volunteer from the
audience choose the card they want, you, by subtle suggestion, will
attempt to make as many people in the audience as possible choose

211
212 Conjuring with Computation

the card you want them to. You will try and draw everyone’s eyes
to the same card. Explain that what you are about to try requires a
deep understanding of the way the brain works, of cognitive psychology.
You explain that you will display on the screen a picture containing
five playing cards and ask everyone to choose one. They can choose any:
perhaps an end one, perhaps the middle, perhaps the ones in between,
perhaps to the left and perhaps to the right. They should let their eyes
settle on one without too much thought and then stare only at it. It
will be their choice. Once they have chosen, they should stand up and
remain standing up.
You show the slide with the five cards, reminding them to go with
the first one their eyes settle on and stick with it, and then stand up.
When all are standing, you switch to a blank slide and tell everyone
they must think about their card and only their card. They must
concentrate purely on it. You will then deal out the cards again, but
you will have removed one card, the one you were trying to suggest
to everyone that they pick. You tell everyone that if they picked the
card you have removed, then they should sit down. Note that with five
cards you would expect one in five of the audience to sit down if you
had no powers of suggestion. Less than that and you would have done
the opposite of what you intended. More and your powers of suggestion
have worked at least to some extent.
You now show a new slide with the four remaining cards, without
the one you were forcing them to choose. Everyone sits down!

How It Works
This is a very simple trick. How it works is that you just replace ALL
the cards with a set of four cards that are similar. Everyone’s card has
gone, so everyone sits down. To make it work, you need to use cards
throughout that are very nondescript: so especially no court cards or
Aces. Make them all 7s, 8s and 9s of similar red and black suits and
swap them for different cards that are also all 7s, 8s and 9s of similar
red and black suits. This ensures that the ones not chosen are not very
noticeable.
 Supernatural Suggestion: Visual Salience 213

The Magical Algorithm

TO PREPARE Supernatural Suggestion:
1. Create three slides, one with five cards in a line, one with four
similar but DIFFERENT cards in a different order. Have a blank
slide in between. The cards should be a mixture of 7s, 8s and 9s
and a mixture of red and black.
TO DO Supernatural Suggestion:
1. Explain that you are going to use your powers of suggestion to
make as many people as possible choose the same card. Talk about
the possibilities based on the positions.
2. Show the slide containing pictures of the five cards.
3. Ask everyone to choose a card, the first their eyes settle on, and
then concentrate on that card alone, standing up when they have
chosen one.
4. When everyone has stood up, replace the slide with a blank one.
5. Explain you are going to remove the card you think most people
will have chosen because of your powers of suggestion. If the card
you remove is the one they chose, they should sit down.
6. Replace the slide with the one containing the four different cards.
7. Explain, as everyone sits down, that you did more than just control
a few people, you controlled everyone and that is the amazing
power of suggestion.

Thinking Computationally (with Psychology

on the Side)
The story behind this trick is all about how you can make a person’s eyes
fix on one thing over another. That is what visual salience is about,
but it just works in a different way than that claimed in the trick. It has
nothing to do with suggestions and everything to do with subtle visual
cues that draw your eye to some things and not others. Of course, the
secret of the trick is a con rather than relying on where your eye is
actually drawn to.
The trick does, however, depend in part on a similar thing of people
NOT noticing the subtle change from a set of similar cards to different
214 Conjuring with Computation

similar cards. It works because some things are more noticeable, more
visually salient, than others. That is why we use 7s, 8s and 9s as they
are less salient than the court cards and Aces. We want them all not
really to be noticed.
Our eyes are drawn to some areas of a scene more than others, and
as we have seen, once our attention is focussed, we miss other things
completely. So what are our eyes drawn to? Our eyes are, for example,
drawn to areas with high-contrast differences. Points in such areas of a
scene are visually salient.
This matters when designing interfaces as it can be used by the
design team to ensure people’s attention is drawn to the right places at
the right time. In doing so, the interaction designer can help ensure that
users of the interface do not miss important things. Having spent 20
minutes choosing things on a shopping website, the last thing you want
is to then not be able to buy them because you can’t find a checkout
button! That is something your eyes should easily and immediately be
drawn to.

PAUSE: Peter W. McOwan

Peter W. McOwan (1962–2019), the co-author of this book before
his death, was an amateur magician but a professional computer
scientist. His research focussed on biologically inspired computing
and specifically understanding the way humans and animals see the
world. He and his team, including Milan Verma and Hamit Soyel,
created artificial intelligence systems based on our understanding of
our vision system. Tools based on his work are now used in a wide
range of areas, including marketing, to predict, for example, what
our eyes are drawn to in ads. He was also passionate about public
engagement in science and a pioneer in using magic to inspire about
computer science.
 Chapter 33

The Numbers Game

Visibility of System State

Conjuring
A member of the audience calls out 10 random numbers. You
write them down, mix them up and spread them out face down,
and a volunteer picks one. Somehow, you are able to tell them the
number they picked.

Computation
It is really important that the design of a computer system prevents
“automation surprises”, where a user loses track of the internal
state of the system. An important interaction design principle to
do this is that of “visibility of system status”. This involves ensuring
that the important parts of the internal state of a computer system
are clearly visible to the user. That means they do not have to
mentally track them, reducing the likelihood of such mistakes.

215
216 Conjuring with Computation

The Trick
Do this as an apparently spontaneous dinner-table trick. Take a piece
of blank paper you happen to have in your pocket and tear it up into
16 pieces. Get someone from the table to call out 16 numbers at random
that you write on the pieces of paper. Shuffle the pieces of paper so they
are thoroughly mixed up. Next, divide them into two piles face down on
the table. Ask someone else to point to a pile allowing you to eliminate
half of the possibilities leaving only eight pieces of paper. Now spread
them out face down on the table and ask a third person to pick one and
look at it without telling anyone what it is. Explain that you can tell
what it was from the micro-expressions that pass over their face as they
look at it. You tell everyone which number you think they have picked.
The person shows everyone the one they took. You got it right.

How It Works
What the audience do not realise is that you don’t write each number
they are calling out, but instead write the first number they said on
every piece of paper.
All the numbers on the cards are the same, so of course, you know
the number on the card they predict. It is only magical because the
audience just assume that you are writing the numbers down as spoken.
For a simpler version, you can skip dividing the paper into two piles.
It just gets more people involved and adds an extra layer of confusion
over what is happening.

The Magical Algorithm

TO DO The Numbers Game:
1. Tear up a piece of paper into 16 equal pieces.
2. REPEAT 16 TIMES.
a. Ask the volunteer to call out a number at random.
b. Write the first number they called out down on one of the
pieces of paper.
c. Place it face down on the table in a pile.
 The Numbers Game: Visibility of System State 217

3. Mix up the pieces of paper ensuring they remain face down

throughout.
4. Split the pieces of paper into two piles.
5. Ask a second volunteer to choose a pile.
6. Spread out the pieces of paper from that pile, keeping them face
down.
7. Ask a third volunteer to choose one piece of paper from the pile
and look at it.
8. Announce that, based on their micro-expressions, you can predict
which number is written on the piece of paper that they chose.
9. Say the number that you wrote on every piece of paper.
10. Have the volunteer show everyone that you were right.

Thinking Computationally (with Psychology

on the Side)
Many card tricks work to some extent because you cannot see exactly
what the cards are. This trick is totally about that. If they could see the
numbers, it would not be magical at all. Good tricks involve surprises!
All tricks involve the magician engineering a system whereby the
audience all make the same mistake at the same time. Part of the mistake
normally involves everyone losing track of the hidden things, such as
face-down cards or here pieces of paper. In this trick, if the numbers
written were all visible so what was on them was clear, then no one
would be surprised at all about the result. It is the lack of visibility of
the numbers on the pieces of paper that makes it seem magical. When,
at the end, the correct number is predicted, you get the sense of delight
that you are after.
Interaction designers are after the opposite effect. Rather than
wanting users to lose track of the internal state by hiding things, you
need to give them enough feedback about the internal state so they do
not get confused in a way that matters. Users must be able to track how
the system is changing internally so they can make appropriate decisions
about what they need to do next.
This leads to a really important interaction design principle con-
cerning the visibility of the system state. We do not, in general, want
218 Conjuring with Computation

people using our computer systems to be surprised to find the system

is in a completely different state to the one they thought. Imagine if
this was the case for a pilot flying a plane. As they come into land,
they suddenly find they are descending really steeply, so steeply they
will hit a hillside, when they thought the plane was set for a shallow
angle of descent. When this kind of thing does happen, it is called an
automation surprise. They are bad because bad things can happen as
a result.
The above plane scenario actually happened with tragic results. An
A320 passenger plane crashed into a hillside when landing at Strasbourg
Airport in 1992. The cause turned out to be that the pilots lost track of
what mode the plane’s autopilot was in. They entered numbers to control
the descent thinking they were entering a shallow angle of descent of
3.3 degrees. In fact, they were telling the plane to descend 3,300 feet
every minute. They had no idea until a few seconds before the plane hit
the hillside. That is an automation surprise at its worst. 87 people died.
Only six survived.
How do you prevent automation surprises? As we use a system,
we mentally track what the system is doing and the internal state we
think it is in. Automation surprises happen because we lose track so
that what we think the state is differs from the actual state. If we make
important internal state constantly visible in a clear way, then this mental
tracking is less important. Instead, we can just watch the changes as they
happen. The Strasbourg plane crash in part happened because the vital
information was not visible in a clear way: the mode of the autopilot
and the current angle of descent were important but not easily visible.
The cockpit design has as a result been changed so that they now are
visible. It would have been better for it not to take a crash for it to be
improved. Of course, if there is too much information visible, then the
information that matters will be lost amongst the rest. A critical part
of interaction design is in working out which information is critical for
given tasks and finding ways to ensure it is clear at a glance at the time
it is needed.
 The Numbers Game: Visibility of System State 219

PAUSE: Don Norman and

Jakob Nielsen
Much technology is far harder to use than it needs to be. Don
Norman and Jakob Nielsen are two of the greatest pioneers of
solutions to this problem. Norman introduced the idea of user-
centred design with Nielsen introducing the related idea of usability
engineering. Both concern making people and their limitations
central in the design of computer systems. More than anyone they
have promoted the idea that with good design and appropriate
engineering methods, software and gadgets can be made far easier to
use . . . if only programmers do take people’s limitations into account.
In Norman’s case, his ideas were based on his eclectic background as
an electronic engineer and psychologist as well as designer. His book
The Design of Everyday Things (1988) is probably one of the most
influential on the topic and certainly a book all computer scientists
should read. Norman and Nielsen have developed practical principles,
guidelines, rules, and evaluation methods that help make software
more usable if used. They joined together to found a consultancy
helping to make technology easier to use.
This page intentionally left blank
 Chapter 34

The Three-Way False Cut

Conceptual Models and Metaphors

Conjuring
This false cut leaves the whole pack unchanged.

Computation
A way to help make interactive systems easy to use is to give them
a clear and simple “conceptual model”. The way they work should
be obvious. One way to do this is to use a design metaphor: you
make the system remind its users of some real-world thing they
already know how to use and then have them interact with the
computer system in a similar way. On the other hand, if you use a
metaphor that does not match the actual design, then the system
will be very hard to use.

The Trick
Place the pack on the table (pile 1). Now cut off the top third or so from
this pile and place these cards to the right. Call this pile 2. Now cut half
of what’s left in pile 1 and place this further to the right of pile 2. We’ll
call this pile 3. All that remains is for you to pause and then reassemble

221
222 Conjuring with Computation

the pack. Place pile 2 onto pile 3 and then take this combined pile and
place it on the cards in pile 1 (see Figure 34.1).
To the audience, this looks like a fair series of cuts, but if you try it
with a pack, you will discover that the pack is in exactly the same order
at the end as it was at the beginning.

The Magical Algorithm

TO DO a Three-Way False Cut:
{No assumptions}
Let there be three positions on the table, which we will call Position 1,
Position 2 and Position 3.
1. Place the pack on the table in Position 1.
2. Cut off approximately the top third to make a second pile in
Position 2.
3. Cut half of the remaining first pile in Position 1 to make a third
pile in Position 3.
4. Pause.
5. Place the second pile from Position 2 onto the third pile in
Position 3.
6. Place the full new third pile from Position 3 onto the first pile in
Position 1.
{The order of the pack is unchanged.}

How It Works
You have just confused the audience. All that takes is to put the piles
back together in a different order to the way they were taken apart!
They are expecting that a series of cuts is fair. Because we put them
back together in a different order to the way the piles were cut, it is easy
to lose track of the detail of what is actually being done which is to put
each third back on top of the one it started off on top of. The pause in
the middle adds a little time misdirection too.
The magician is deliberately projecting an idea of how the world
works that is wrong. Everyone who has played cards knows without
thinking that if you cut a pack of cards, they end up in a different order.
The Three-Way False Cut: Conceptual Models and Metaphors 223

Figure 34.1. The steps of a three-way false cut.

224 Conjuring with Computation

It is a standard way to ensure the cards dealt are unpredictable. Each

part of the cut follows the pattern of a real cut. You have removed part
of the deck of cards and then put part of the deck on top. By cutting
the pack, you set up an expectation, and rather than concentrating on
the detail of exactly what is happening, our brain uses this model of the
world to predict what will happen.

Thinking Computationally (with Psychology

on the Side)
When designing human–computer interaction, we, as in so many
situations, need to do the opposite of the magician. Just as with the
magic trick, we find it very hard to track the internal state of a program
as it changes, even when the change is simple. The design of an interface
as presented to us (also known as the system image) leads us to create
an expectation or a conceptual model of how it works. In the trick,
the system image presented is of fair cuts. We then do not track the
detail, but use this simplified expectation, to work out what we think
will happen. If the model we create in our heads (called the user model)
is not the same as the one the designers actually used, so not how the
design actually works (called the design model), then we will make
mistakes about what the state of the system is.
In this trick, we set up a system image so that the viewer creates
a user model of fair cuts being made. In reality, however, the design
model is that we are splitting the pack and then putting it back again
in the same order, not a different one.
The programmer has a really detailed idea of how the system they
are programming works so it is often very hard for them to see how others
could be confused. With their detailed knowledge, it is really obvious to
them how it works and how tasks are to be done. Their personal user
model is the same as the design model and is not created from the
system image.
The user interface of the program (projecting the system image)
only projects an imperfect version of what is happening. Unlike the
programmer, the user only has this to go on. Based on their past
experience and what they can see from the interface, a user forms a
 The Three-Way False Cut: Conceptual Models and Metaphors 225

mental model of how they expect it to work and what they have to do.
Those expectations may or may not match the actual system. If it is an
accurate version of the system behaviour, then it will help the person
use the system with ease. However, if it is not accurate, it leads to
misconceptions of what to do when using the system.
For example, if you have seen lots of hand dryers in public toilets
that are activated by a large button, then when you see a new hand
dryer, you will form a user model that includes the expectation to press
a button. If in the place of the button you see something button-like,
then this will reinforce that user model. You are likely to press it without
thinking. This is likely even if it is actually just a sign in a box saying
“DO NOT PRESS: place hands under dryer to activate”.
Magicians set up system images of the magical system explicitly so
that users form the wrong mental model of what is happening. That is
what is happening in this cut. Everyone has seen people cutting packs
of cards before, and they know that it is one way to mix up a pack and
bring random cards to the top. The first cuts look like that is what is
happening, and rebuilding the pack seems to fit what happens with a
sequence of fair cuts. The magician’s presented system image makes the
audience form a user model that the pack is being randomised a little.
They do not follow the detail but instead assume they know what is
happening.
Programmers and interface designers need to do the opposite. Bad
designs work like a trick making the user form a faulty mental model,
and good designs help them form a correct one: so hand dryers that
are triggered by sensors should have nothing that looks like a button
that will reinforce the wrong mental model. Better still they should look
completely different and ideally have some very obvious visual clue that
you put your hands in the right place. The same considerations apply to
interfaces for more complex software.

Design metaphors
One way to help users quickly form the right mental model is to make
the controls look like some familiar real-world thing. For example, a paint
program might have a set of colours to choose from that look like they
226 Conjuring with Computation

are laid out on a painter’s pallet. To change colour, the user moves their
virtual brush to “dip” it in the paint on the pallet.
This is called a design metaphor, and it helps the user understand
the way the system works. The actions needed to achieve a task,
like choosing a colour, match those in the real world when dipping a
paintbrush in paint. The interaction design principle to remember is
that if it looks like something familiar, then it should behave like it.
If it looks like a painter’s pallet, then it should behave like a painter’s
pallet . . . or it will cause confusion.
To take another example, the metaphor of a dustbin is used on many
computer desktops so you naturally understand that files you put there
are “rubbish” so intended to be deleted. However, they are not deleted
immediately. That happens only later when we empty the trash. We
have that expectation, and so if that is the way the system works, we
naturally understand the right thing to do. There is no actual dustbin
in a computer, of course, it is just a metaphor. If instead dropping
something in the virtual bin deleted it immediately, then people would
likely often make mistakes. That would need a different metaphor to
help: a shredder metaphor perhaps. When we shred things, we expect
them to be gone for good immediately. It is what we see (the icons), the
language used, and the feedback we see that sets up the expectation.
Similarly, in the trick, the magician sets up the metaphor of doing
a cut, and so we use that as our conceptual model watching the trick.
We use that metaphor to judge what happens. What is done is close
enough to our expectations that no alarms are raised. We know what is
happening. The design model, what is actually happening, is, however,
different. As the magician hopes, our expectations mean that we make
a mistake and do not realise the pack is unchanged.
When we design and evaluate computational systems, we don’t want
people getting confused, so need to ensure we take expectations like this
into account. We need to design in a way so that the interface leads to
the user’s conceptual model matching the design model. We also need
to check whether we have succeeded, evaluating the design with real
users to ensure that they do understand the metaphor and it sets up the
correct expectations.
 The Three-Way False Cut: Conceptual Models and Metaphors 227

The magician uses metaphors to confuse us so we miss what is really

happening. Good interaction designers use them in the opposite way to
help avoid confusion and so make interfaces easy to use.

Interlude: Jay Ose

This false cut is credited to magician Jay Ose (1911–1967). He was
amazing at close-up card tricks and was the advisor on various films
involving card sharps and con artists, including Steve McQueen’s
Cincinnati Kid. He appeared in a short documentary about the
making of the film demonstrating his card-handling skills. He also
appeared in other films and TV shows through the 1960s, including
an episode of Mission: Impossible.
This page intentionally left blank
 Chapter 35

The Joker in the Pack

Memory

Conjuring
This trick is about finding troublemakers (i.e., jokers) in a
classroom and getting them to the front of the class. A volunteer
thinks of a number, deals out that many cards, and amazingly they
have brought the only joker in the pack to the top.

Computation
No one has a perfect memory. We all forget things, especially
short-term things, as our limited working memory can quickly be
filled. Forgetting something we know we have to do in the future
is a particularly common mistake, especially in stressful situations.
Interaction designers need to keep our memory limitations in mind,
ensuring the interface only requires us to remember things if they
can’t be avoided.

The Trick
Tell the audience that teachers must be able to spot troublemakers: the
jokers of the class. You will show them a test for prospective teachers

229
230 Conjuring with Computation

to see if they can do the necessary crowd control. The person being
tested will use their natural talent at detecting problems before they
happen. They will try and find a joker in the pack. A volunteer thinks of
a number, say between 2 and 8, and announces what it is. You shuffle
the pack and then show them what you want them to do. You count
out their chosen number of cards aloud 1, 2, 3, . . . from the top of the
pack of cards onto the table, face down. When their stated number of
cards have been dealt, turn over the last card dealt and point out what
it is. Place the cards back on top of the pack and hand the whole pack
to them. Take some time to remind them that they are looking for the
joker, which stands for a naughty student who is about to cause trouble.
They must get them to the front of the class (so the top of the pack),
where you can see them before they have a chance to make mischief.
While doing this, you shuffle the pack.
They now do as you instructed, counting out the cards, 1, 2, 3, . . .
up to their number. Before they turn over the top card, take back the
rest of the cards that weren’t dealt out and spread them face up on the
table, pointing out that there are no jokers. Pull out some single cards
and say aloud what they are, to emphasise they are not jokers. Remind
everyone that you shuffled the pack. Now have them check to see if
they found the joker by turning over the top card of the pile they dealt.
Amazingly, the card they turn over is the joker!
Despite it being the only one in the pack, they have brought the
troublemaker to the front with no trouble at all. They are a natural at
crowd control and a natural for the classroom.

How It Works
To make the trick work, before the performance, you simply place a joker
on top of a shuffled pack. Each time you shuffle the cards during the
trick, you do a fake shuffle that leaves the top cards unchanged while
appearing to mix the cards up (e.g., a False Top Overhand Shuffle: see
Chapter 18).
When you count the cards onto the table, do it casually so that they
think nothing of it: you are just showing them what to do.
 The Joker in the Pack: Memory 231

The joker starts at the top of the pack. The fake shuffle leaves
it there. Suppose the person chooses the number 4, then in your
demonstration, you count 4 cards onto the table face down. The joker
is now at the bottom of these cards, so in this case, it is the fourth from
the top. You place the cards back on the top of the pack so it is now
the fourth card in the whole pack. A similar thing happens whatever
number is chosen (see Figure 35.1). The fake shuffle again leaves the

Figure 35.1. If the volunteer picks number 3, then the joker ends up in the third
position.
232 Conjuring with Computation

top cards alone, so given they were asked to pick a small number,
the shuffle should leave their card in the same position from the top.
You hand the pack to them, and they now do exactly as you did.
However, now counting four cards out actually brings the fourth card
(the joker) back to the top. It is the card they later turn over. Whatever
number they choose, they exactly undo what you did.
Your extra patter between you dealing and them dealing and also
between them dealing and then finally revealing the joker are important.
It takes the audience’s attention away from the cards dealt and from
the pack. The time delay it causes also helps, as this, combined with
all the extra things you are doing, increases the chance that they forget
exactly the detail of what you did at the start. This is time misdirection.
If you are not confident doing fake shuffles, then leave them out: the
trick still works as long as you have time misdirection, it is just slightly
less powerful.

The Magical Algorithm

TO PREPARE the Joker in the Pack:
1. Place a joker face down on the top of a shuffled pack.
TO DO the Joker in the Pack:
1. Explain this is a test for whether someone can find troublemakers.
2. Ask a volunteer to pick a small number, between 2 and 8, say
and tell everyone what it is.
3. DO a False Top Overhand Shuffle, preserving the top 8 cards.
4. Show the volunteer what they must do by dealing their chosen
number of cards onto the table.
5. Turn over the last card dealt saying what it is and noting they
are trying to find the joker and bring it to the top then put the
dealt cards back on top of the pack face down.
6. DO a False Top Overhand Shuffle, preserving the top 8 cards.
7. Take time talking about keeping troublemakers near you.
8. Give the volunteer the pack.
9. Have them count out their chosen number of cards.
10. Turn over the remaining pack and show that there are no jokers
there.
 The Joker in the Pack: Memory 233

11. Have them turn over the top card of the dealt pack.
12. Point out they brought the joker to the top of a shuffled pack so
are a natural at spotting troublemakers.

Thinking Computationally (with Psychology

on the Side)
This trick relies on our fallible memory. The audience at the time do not
understand the importance of you counting out the cards. Your intention
is that by the time they can make sense of what has happened, they
have completely forgotten you even did it. This is done by a combination
of time misdirection and giving people other things to think about.
As an interface designer, you have to take into account human
fallibility, and there are many ways we are fallible and so make mistakes.
As we have already seen, we have limited attention. We also have very
limited working memory: that is only a small and fixed amount of short-
term memory. Many kinds of mistakes that we make are linked to the
fallibility of our short-term memory.
Take car lights, for example. It is quite common for drivers to get out
of their car having forgotten to turn the lights off. It is easy to end up
flattening your car battery as a result of such forgetfulness. Why is this
so common? They switched the lights on so knew they were on. They
knew they needed to switch them off before leaving the car. Some may
even have made the mistake before with bad consequences like a flat
battery. Still, they forget. They were relying on their working memory to
remember, but in the meantime, there was a long delay while they did all
the driving. Driving involves concentrating on lots of other things that
fill up working memory. Eventually, they arrived at their destination, so
had achieved what they were trying to do, and now had new things to
think about: greeting their family or getting on with work, for example.
The little step of switching off the lights went out the window!
This kind of human error is called a prospective memory error:
knowing you have something to do in the future but then forgetting
about it when the time comes. In this case, what was forgotten was
a tidying up task that could only be done after the person’s goal had
been achieved. You cannot turn off your headlights until after you have
234 Conjuring with Computation

arrived! Ever gone to find something in your bedroom and forgotten to

turn the light off when you left? Your goal was to get the thing you
were after. Turning off the light is just a tidy-up task. Experiments have
shown that that particular kind of prospective memory error happens
when a person’s working memory is filled. The more short-term things
you have to remember the more likely you will make this kind of mistake.
No one is immune. Everyone can get it wrong if given too many things
to remember, though some people have a bigger capacity than others.
People make this kind of mistake in all sorts of situations, not just
with lights, and including using computer systems. Rather than human
error, they should be thought of as the fault of the designer of the
system. Some simple design changes can make them far less likely,
and sometimes impossible. A really simple example is in the design of
cashpoints. Cashpoints that give money before returning the card are
setting up people to make this kind of mistake. The person’s goal is to
get money, so retrieving the card becomes a tidy-up task to be done
afterwards. If the machine requires the card to be retrieved first, then
the person is less likely to forget as it would mean leaving without the
money they came for too. Forgetting the money is possible but less likely
as it is the reason for using the machine.
In some similar situations, like paying with a card at a supermarket
or a garage, the problem can be eliminated if the person only scans the
card rather than giving it up. Then there is no longer a possibility at all
of forgetting to take it back as it is never handed over in the first place.
There are a range of ways that memory can lead to problems with
poor design. Another common source of problems occurs when a person
is interrupted in the middle of doing a task. Then, when they try to
continue, they are likely to either redo the last step they did (forgetting
they did it) or skip a step because they think they did the last step when
they only thought about doing it. Interface designers can help avoid such
problems by ensuring that in a sequence of steps the interface clearly
shows what step the person is actually on.
 The Joker in the Pack: Memory 235

PAUSE: The Xerox PARC Team

One of the biggest steps forward in making computers easy to
use was the Graphical User Interface (GUI). It makes a massive
difference for lots of reasons but part of it is because it removes
the need for people to remember obscure commands to get things
done. Instead of typing commands, they can just click on things
they recognise and, e.g., drag them about (“drag and drop”).
A big team from Xerox PARC were responsible for creating the
first GUI. It included Alan Kay, Larry Tesler (1945–2020), David
Smith, Clarence Ellis (1943–2014) and Dan Ingalls. They all made
significant contributions and went on to do other amazing things
too. For example, Clarence Ellis went on to work on early versions of
shared documents that allow lots of people to edit them at once (as
is now taken for granted with Google Docs), and David Smith came
up with the idea of icons (pictures to click on that run programs) and
a way to program robots by demonstrating actions, Larry Tesler had
the idea of cut and paste as a way of copying things and later helped
improve shopping websites, and Dan Ingalls came up with pop-up
menus. Alan Kay, as well as working on the first GUI, also made
massive contributions to the design of object-oriented programming
languages: where programs are thought of as a collection of objects
that send messages to each other, as opposed to being like recipes
to be followed.
This page intentionally left blank
 Chapter 36

Are You Psychic?

User Experience and
Engineering Delight

Conjuring
You test the psychic abilities of a volunteer and show that they
have great power. They successfully match pairs of cards that had
been mixed up, even though those cards were face down the whole
time.

Computation
Ideally, you do not want your software just to be usable, you want
the user to have a great experience. To give a good experience,
you need to engineer that experience.

The Trick
First, you need two sets of matching cards. You can do it with any
cards that can be paired. Zener cards, which you can buy from magic
shops, are ideal as they are the cards that scientists have used to test for

237
238 Conjuring with Computation

psychic powers (or at least to prove people aren’t psychic). They have
different patterns: a circle, a cross, wavy lines, a square and a star. You
can, however, do the trick with happy family cards or playing cards if you
have two identical packs. We will do the latter here. You also need four
“magical tokens”. Anything that looks a bit magical will do like gems,
toy crystals, tokens or just coins.
Beforehand, take five cards: if using normal cards, an Ace, a King,
a Queen, a Jack and a 10, and shuffle them. Then take another set of
identical cards and put them in the same order. Place one pile of five
cards on top of the other. So, for example, if the first five end up in the
order King, 10, Jack, Ace and Queen, then the whole sequence is King,
10, Jack, Ace, Queen, King, 10, Jack, Ace and Queen.
Do a Cyclic False Cut (Chapter 20) by showing a volunteer from
the audience the cards face up, fanned out in your hands so they can
see there is a mix of different cards. To then mix up the cards, turn the
cards face down and allow the volunteer to point to a card. Split the
pack at that point, putting the top part to the bottom. Keep doing this
until they are happy the cards are mixed. If doing the trick as close-up
magic with people gathered around the table, you can have different
people touch the cards.
Next deal the top five cards, one at a time, into a pile on the table.
Place the remaining undealt cards into a second pile beside them.
Tell the volunteer that, now the cards are mixed up, they must try to
end with two cards that match without seeing any of the cards. This will
prove whether or not they have psychic or magical powers. Explain that
they have four magic tokens to channel their powers. Each will stand for
a move. A move involves taking the top card from one of the piles and
placing it at the bottom of the same pile. They can put all four tokens
on one pile, and none on the other; two on each; or three on one and
one on the other. It is their choice, remembering that their aim is to
be left with two matching cards. You will do the number of moves on
each pile that corresponds to the number of tokens there. The volunteer
will therefore be the one deciding what happens by placing the magic
tokens on the piles. Once the tokens have been placed, you make that
many “moves” on each pile. A “move” just involves moving a card from
the top of the pile to the bottom. So if a pile has three tokens placed
 Are You Psychic? User Experience and Engineering Delight 239

on it, you move three cards from top to bottom, one at a time (see
Figure 36.1).
Next, remove the top card from each pile and place them aside in
a pile together. Place one of the magic tokens on top of them to show
they are discarded. It does not matter what they are; it is the cards left
at the end that must match.
Now there are four cards in each pile. Give the volunteer the three
remaining tokens to repeat the process. Once the moves are done,
remove the top two cards from the piles again, placing a token on them
as before. There are now three cards left in each pile and two tokens. Do
it all again, again removing the top card from both piles. This leaves two
cards in each pile and a single magic token. This is their final chance to
get it right. One move remains, and one card can make all the difference.
They choose where to put the token, you make the last move, and the
top two cards from each pile are discarded.
Now it’s time to reveal the final two single cards left on the table.
No one has seen them through the whole process, they were mixed at
the start, and the number of cards moved was completely chosen by
the volunteer. Ask the volunteer again if they think they have psychic
powers. Tell the audience to cheer if the cards do match, but give a loud
groan if they don’t. Turn one card over, asking if they think the other
will be the same, then turn the final card over. They match!
The volunteer chose freely which cards to eliminate, so it was their
secret psychic powers that came through to ensure a match at the end.

Figure 36.1. Tokens are spread between the two piles. The number of tokens is
always one less than the cards in each pile (here five cards, so four tokens). The same
number of cards are moved from top to bottom as tokens are placed on a pile.
240 Conjuring with Computation

Give their jaw time to drop, then say that once you have found someone
with magical powers, it is worth checking how powerful they are. One
pair at a time, dramatically reveal that all the pairs of cards they removed
match too!
Point out that they should stop denying their powers, go and buy
a superhero costume, and start using their powers for the good of
humanity!

The Magical Algorithm

TO PREPARE Are You Psychic?
1. Take the Ace, King, Queen, Jack and 10 of the same suit from
two identical packs of cards.
2. Shuffle the five cards from the first pack.
3. Put the five cards from the other pack in the same order.
4. Place one set of five cards on top of the others.
5. Place four “magic tokens” (e.g., gems) on the table.
TO DO Are You Psychic?
1. Spread the cards fanned face up to show that they are a mixture
of cards.
2. Ask a volunteer to come to the front.
3. DO THE FOLLOWING UNTIL the volunteer is happy the cards
are sufficiently mixed.
a. DO a Cyclic False Cut.
4. Deal the top five cards onto the table, one at a time.
5. Place the other five cards together in a pile on the table next to
them.
6. Explain that the volunteer must, without seeing the cards, use
whatever psychic powers they have to try to be left with two
matching cards at the end.
7. Give them the four “magic” tokens.
8. DO THE FOLLOWING FOUR TIMES.
a. Ask the volunteer to place each token on one pile or the other
to direct what you do.
b. Move the number of cards equal to the number of tokens on
each pile, one at a time, from top to bottom of that pile.
 Are You Psychic? User Experience and Engineering Delight 241

c. Remove the top card from each pile and place them aside
together with one of the tokens on top.
d. Give any remaining tokens back to the volunteer.
9. Turn over the two cards that remain to reveal that the volunteer
does have psychic powers! Take the applause.
10. Point out that once you have found someone with psychic
powers, it is worth seeing how powerful those powers are.
11. DO THE FOLLOWING FOUR TIMES.
a. Turn over one of the remaining pairs put aside during the trick
to show that they match.
12. Suggest the volunteer should stop denying their powers, buy a
superhero costume, and start using their powers for the good of
humanity!

How It Works
The trick works because of a combination of the starting order, the way
the cards are then mixed up, the way they are placed on the table at the
start, and then the fact that there is always one less token to place than
the number of cards remaining in each pile.
First, the cards start off in a pile with five cards followed by the
same five cards. The way the cards are mixed is then via a Cyclic False
Cut (see Chapter 20). Each time you cut the pack at the point of their
choosing, you are just rotating the order of the cards. At all times, there
are still five cards followed by the same five cards as at the start. The
sequence just starts with a different card. It is the cyclic order that
matters though, and that is unchanged.
Next, you count the first five cards out onto the table. Critically,
this reverses their order. The remaining cards are placed straight onto
the table. Their order is unchanged. That means the two piles hold the
same five cards but in opposite orders.
We finally always have one magic token less than the cards in
each pile. It turns out that when the cards are organised as they are
(an identical sequence except with the order reversed), how ever you
share the tokens between the piles, you will always end up with a pair of
matching cards on the top of the piles. That pair of matching cards is the
242 Conjuring with Computation

pair that is discarded. We will leave the details of why this happens to
the next chapter (see if you can work it out in the meantime). Here, the
important point is that on every round, you are actually guaranteed to
discard two cards that match. That means the final pair must too, and
it allows you to take the audience by surprise with your twist of a final
reveal.

Thinking Computationally (with Psychology

on the Side)
This trick shows that you can engineer a system in a way that increases
the delight that results. It uses a fairly simple principle. Set up an
expectation. Meet that expectation. Then go further and do something
unexpected (but rewarding).
Here the audience almost certainly expected the final two cards to
match. It wouldn’t be magic if they didn’t. While that is entertaining
because it is hard to see how it could have happened, it is not
unexpected. However, few will have expected all the cards to match
as that seems very unlikely and you did not imply at all that it would
happen, so that is a surprise. Given matching two cards by chance
seemed unlikely, matching all of them is by comparison astounding and
so leads to greater delight.
Just as we can build that idea into the experience of a magic trick, we
can build it into the experience of using software. Engineering delight, in
this way, is a part of a wider aspect of design and evaluation of software:
that of user experience. User experience is about provoking the right
emotions in your users.
Why does this matter? Well, if, for example, the software is a
shopping website, it does if you wish to build customer loyalty. Suppose
a disabled person has bought a concert ticket online but finds that they
have to change the date. They hope they can do it, though worry they
won’t, but they assume it won’t be easy and expect they will have to
pay a charge and will have trouble getting a new disabled access seat.
That has been their experience in the past. They go to the website and
immediately see a clear way to change tickets. The system has very good
 Are You Psychic? User Experience and Engineering Delight 243

usability in that it is easy to use. That is providing a good experience.

However, it is also manipulating their emotions. They were worried (an
emotion) before in case they couldn’t work out how, or that it might
be impossible, but now are relieved. Better still, and surprisingly, it tells
them it has automatically booked a disabled seat on the new date rather
than them having to even mention it. Finally, they find not only is it
all easy to do but also are explicitly then told that there is no charge.
They are delighted and rave about the ticket site to others, and vow to
always use that site in the future.
One of the most powerful ways of delivering delight is to design
something that people hadn’t realised they needed, but suddenly do so.
For the delight to last, it must actually fulfil a need they really did
have, though. For example, I bought a soap dispenser on the begging
of my then 5-year old who loved the cartoon on the label. After we
had bought it, we found out that it wasn’t a normal soap dispenser but
an interactive one. It giggled every time you pressed the button to get
soap. That in itself was delightful the first time. What really delighted me
long term though was that we no longer had to nag about handwashing
as my 5-year old now actively wanted to wash his hands. It delivered
what I expected, but surprisingly, then went much further, and delivered
something I needed without realising it.
Finding such needs takes a lot of experience. The important thing,
though, is to know the people you are designing for well. You need to
understand their unfulfilled needs. Rather than having a preconceived
idea of what they need, you must understand their lives. For example, if
you are tasked to come up with a product for families, then by talking
to parents, you might discover that getting kids to wash their hands is
a problem. That is the need so becomes the design goal. Design a good
solution, and you have a big advantage over competitors, especially if
those who buy it are delighted when they realise their tricky problem is
solved.
244 Conjuring with Computation

Interlude: Fay Presto

Self-working tricks like this one are often suitable for close-up magic
where, rather than the magic taking place on a stage, distant
from the audience, it takes place anywhere, including on the street.
Spectators gather around and watch from all angles. If spectators
are expected to be watching from a distance but instead are
allowed close up, it improves their experience of the magic. Involving
everyone in a trick is also a way to increase everyone’s pleasure
in watching it. Fay Presto is credited with starting a revolution in
this kind of close-up magic by doing tricks in bars and restaurants,
rather than on a theatre stage. She also held a residency in a London
brasserie. Her most famous trick, passing a bottle through a table,
was voted one of the top 50 tricks of all time.
 Part 9

Evaluation and Logical Thinking II

We now look at more advanced logical thinking about tricks and

programs including real-world uses of algebra.
This page intentionally left blank
 Chapter 37

Are You Psychic? (Continued)

Loop Invariants and Inductive Proof

Conjuring
We return to the trick that proves an audience member has
superhero powers and prove here it will always work.

Computation
When proving that loops do the correct thing, the idea of a
loop invariant provides the basis of logical reasoning: reasoning
by induction.

The Trick
Rather than showing you a new trick, we return to the previous one
(“Are You Psychic?”, Chapter 36) and look at why it always works.
The trick is made up of several parts, but the core part is the repeated
placing of tokens. Below, we give that key part of the algorithm that we
now focus on. The first part of the trick sets up a precondition: on the
table are two piles of five cards each, face down, with the same cards in
each but in the opposite order. At the end of this part of the algorithm,
it guarantees the postcondition that all pairs of cards match.

247
248 Conjuring with Computation

The Magical Algorithm

{On the table are two piles of five cards each, face down, with the same
cards in each but in the opposite order.}
1. Give the volunteer the four “magic” tokens.
2. DO THE FOLLOWING FOUR TIMES.
a. Ask the volunteer to place each token on one pile or the other
to direct what you do.
b. Move the number of cards equal to the number of tokens on
each pile, one at a time, from top to bottom of that pile.
c. Remove the top card from each pile and place them aside with
one of the tokens on top.
d. Give any remaining tokens back to the volunteer.
{Every pair of cards match.}

How It Works
It works because each round of placing tokens, moving cards and
discarding the top two as a pair guarantees that the pair discarded
matches. We have said this is so but how can we be sure it always works?
We need to do some logical thinking. It will be based on something called
a loop invariant and a form of reasoning mathematicians call inductive
proof.

Thinking Computationally
Loop invariants provide a way to reason about the correctness of loops.
They are properties that must always be true at the same point in
the instructions of a loop, usually the start. The property holds every
time the loop’s instructions are executed and get back to that point.
The situation is changed by the instructions in the loop so that the
property temporarily ceases to hold, but is then restored so the property
becomes true again. We use this property as a way to prove that at the
end a related desired property holds. In our trick, that final property is
that all pairs match. This property is the one the loop is being used to
 Are You Psychic? (Continued): Loop Invariants and Inductive Proof 249

guarantee. It needs a slightly more complicated loop invariant property

to guarantee it.
The idea behind inductive reasoning in this context is that if we
can prove three things about a loop invariant, then we can be sure the
algorithm works. We must prove three assertions:

A. The loop invariant property is true at the start, and

B. IF it is true at the start of the loop body, THEN it will always be
true again after following the loop body instructions once, and
C. IF it is true at the end of the loop, THEN this implies that the
final postcondition holds so the algorithm works.

This is a kind of proof called mathematical induction, and it is the

kind of logical reasoning needed to prove that an algorithm containing
repetition always works.
Inductive reasoning works a bit like a paper chain, each link in the
chain is identical, just as the loop invariant is always the same. On its
own, it does not get you very far, but as you chain lots together, you
end up at the other side of the room, where you were trying to get
to, which can be any distance away. The first link in the chain is like
the property being initially true (assertion A above). This is a bit like
pinning the chain to the wall. Add a link by applying the reasoning that
it remains true (assertion B) and you know the property is true after one
time around the loop, one link further. Apply assertion B again, adding
another link and you know it is true after twice around the loop. Keep
applying assertion B, adding more links, as many times as needed, and
you know it is always true: you span the room! Assertion C just checks
that you have arrived with the last link applied at the place you want to
be (pinning the chain to the opposite wall!). It shows that the algorithm
works and the magic effect has been created.

The loop invariant for our trick

If we are going to use this idea to prove our magic trick always works,
we first need to decide what its loop invariant is. What is always true for
our trick when we return to the start of the loop? The actual number
of cards changes as does the number of tokens, so they are not part of
250 Conjuring with Computation

the invariant. However, we know several facts that do stay the same, so
can form our loop invariant:

(I) Both piles contain the same number of cards.

(II) The number of tokens is one less than the number of cards in
each pile.
(III) All (so far) discarded pairs of cards match.
(IV) The cards in each pile are identical except for being in the
opposite order.

Proving that the loop invariant is true before the loop

The first part of our logical reasoning (proving assertion A) involves
showing that this property is true when we first enter the loop. This can
easily be seen just by looking at each fact (I)–(IV) in turn.
We started with 10 cards and split them into two equal piles, so
each pile contains five cards. That means (I) is true.
We gave the volunteer four tokens. That is one less than the number
of cards which we just saw was five cards, so (II) is also true.
There are no discarded pairs at the start, so that means (III) is true
automatically.
Fact (IV) is also guaranteed at the start because of the setup and
false shuffle. The cards started with five cards in some order followed by
the same five cards in the same order. The false cyclic cutting left that
property the same throughout. We then dealt the first five cards onto
the table one at a time so that the first one ended up at the bottom, the
second on top of it, and so on. That reversed their order. Finally, we put
the other five cards down without reversing their order. Consequently,
as required, fact (IV) is true just before the loop is entered.
All of the separate facts we have included in our invariants (I)–(IV)
are set up to be initially true, so our invariant holds at the start, just
before we allow the volunteer to place tokens for the first time. The first
link in our chain of reasoning is in place.

Proving that the loop invariant remains true

The next step (proving assertion B) is to make sure the way the tokens
are placed and the way we move the cards ensures our loop invariant
 Are You Psychic? (Continued): Loop Invariants and Inductive Proof 251

is ultimately left alone (even if temporarily disturbed) by the steps we

repeat, even as cards and tokens are discarded. Again let’s look at each
fact in turn.
At the end of each round, we have discarded one card from each
pile, so while they both have less cards, both piles still contain the same
number of cards as each other given they did initially. Our first fact (I)
is still true.
Is (II) also still true? We have discarded a card from each pile, but
we have also discarded a token with it. That means that (II) the number
of tokens is still one less than the number of cards in each pile, as long
as it was at the start of the loop.
Fact (III) is a little more tricky and is the crux of the trick. It again
relies on there always being one less token than cards in each pile. The
volunteer has a free choice of placing the tokens, so there are several
cases to consider. Let’s think about just one case first and then see if
we can work out what is going on more generally.
If all the tokens are placed on one pile, then the other pile is left
alone. Its top card will be the one that is discarded. The matching card
is at the bottom of the other pile. However, as there is one less token
than cards and all the other tokens are on the other pile, we will rotate
all but the last to the bottom, leaving that bottom card now on the top
to be discarded. The two cards that will be discarded match as required.
Now let us instead suppose that all but one token is placed on one
pile and the last token on the other pile. From the pile with one card,
we will move one card to the bottom and that pile’s second card will
be discarded. It matches the card second from the bottom of the other
pile. However, we have two less tokens than cards on the other pile, so
we will cycle all but two cards to the bottom, leaving the second from
the bottom one now at the top to be discarded. Again the discarded
pair match.
The same reasoning applies in each other case we consider in turn.
We move one more place down the first pile, but also one less place down
the other, so always arriving at the same two matching cards, because
of the reverse order property.
This reasoning follows how ever many cards, and no matter how
the tokens are placed. It works as long as there is one less token than
cards and the cards remain in reverse order. Our other facts mean both
252 Conjuring with Computation

are true. In fact, what we have just presented is actually an informal

version of another inductive proof. Perhaps you can spell it out more
formally in terms of the three conditions once you have read the rest of
the proof.
For (IV), we need to be sure the cards remain in the opposite order
to the other pile. First, we know the moves we do, moving a card from
top to bottom, keep the cyclic order. All it does is change the starting
point in a given sequence so that a different card is at the top and
bottom of each pile. Do we leave the same card at the top of one pile
and the bottom of the other by the number of moves done though?
This relies on similar reasoning to the above. Again we will consider
the case where all tokens are placed on one pile first. One pile is left
alone. In the other, the cards are cycled so that the bottom card comes
to the top. This means the new bottom card now matches the second
card of the other pile. However, we then remove both top cards, leaving
the bottom card of one matching the top card of the other, but with
otherwise the orders the same, as required. The same happens in each
other way the tokens are placed because of similar reasoning to the
above as we always cycle the top cards to match.
Overall, we have just shown that if our loop invariant property was
true at the start of a round of placing tokens, it will still be true when
we next get to the start of a new round of placing tokens.

Proving that the loop invariant being true at the end

implies our magic trick works
Finally, we need to prove that at the end of the trick, if the loop invariant
holds, then that means our magical effect holds: that all the pairs match.
Part of the invariant that we have shown holds throughout is that the
pairs discarded so far do match. At the end, we have discarded all but
the final pair so those are the only ones left to check. However, as we
started with two sets of the same five cards, if all the others match,
these last two must be the same as there is nothing else left that they
can be. Our invariant guarantees that all the pairs match at the end.
Our chain of reasoning has taken us to the place, the conclusion, that
we wanted. We have proved the trick always works. Indeed, our proof
 Are You Psychic? (Continued): Loop Invariants and Inductive Proof 253

does not rely at all on the actual number of cards involved, so the trick
will work technically (and the proof applies) whatever the number of
cards we start with.
Oddly, a property that stays the same is the key to understanding
how doing repetitive computation changes things in ways that make an
algorithm always work. To really be able to code with loops and be
sure the code is correct, you need to be instinctively doing this kind of
reasoning every time you write a loop.

Interlude: Adelaide Herrmann

The false stereotype of the magician is, like that of the computer
scientist, male. In both cases, many women have led the field even
in the early days of computing and stage magic. Adelaide Herrmann
(1853–1932) started out as a magician working with her husband
Alexander, though was far more than just the stereotype pretty
assistant. Her input was integral to many of the tricks, and she
was given equal billing. When her husband died, she continued
doing magic shows on her own for another quarter of a century,
billed as “The Queen of Magic”. She was an active role model who
encouraged other women to take up magic. Her biggest, Noah’s ark-
linked, illusion involved a vast parade of animals magically appearing
out of an ark filled with water to represent the flood. She is also
famous for doing a very dangerous trick that had been the preserve
of men, where she stood in front of a firing squad, catching their
bullets on a plate.
This page intentionally left blank
 Chapter 38

The Red–Black Mind Meld

Abstraction, Algebra and Proof

Conjuring
You meld your mind with that of a volunteer so that you can
predict a fact about the way the red and black cards fall unseen.

Computation
Rigorous argument is one way to convince ourselves that a trick
or program works, but we can go a step further and use algebra
to do this. We describe the trick or program with maths, do some
algebra on the resulting “mathematical model” and prove desirable
properties hold. This approach is a vital way to check safety-critical
software is correct.

The Trick
Ask for a volunteer to take part in the mind meld. Shuffle a full pack
of 52 cards using a normal overhand shuffle (Chapter 15). Deal the
cards into two equal piles and hand one pile face down to the volunteer,

255
256 Conjuring with Computation

keeping the other. The volunteer should not look at their cards, but you
turn your pile face up. Now say you are beginning the mind meld from
you to the volunteer. You will control their actions even as they think
they are making their own decisions.
Select, at random, some BLACK cards from your pile and place
them in a face-up pile in front of you, saying how many BLACK cards
you have chosen. Ask the volunteer to select the same number of cards
(without looking!) from the top, middle or bottom of their pile. They
should do whichever feels right as it will be the mind-meld working!
They should place their chosen cards, face down, in a new pile in front
of your BLACK pile. Explain this pile will be influenced by your BLACK
pile.
Now repeat the process but take a random number of RED cards
from your hand and place them in a new face-up pile next to your black
pile. Tell the volunteer how many RED cards you have chosen. Again ask
the volunteer to blindly select the same number of cards from anywhere
in their pack. They should place them in front of your RED pile. Explain
that this new pile will be influenced by your RED pile.
Repeat the process, alternating between choosing BLACK cards and
choosing RED cards until you have run out of cards. If done correctly,
the volunteer will run out of cards at the same time.
There is now, on the table (see Figure 38.1):

• Pile 1: A face-up RED pile.

• Pile 2: A face-up BLACK pile.
• Pile 3: A face-down pile in front of the RED pile containing the
same number of face-down cards the volunteer chose while under
the mind meld.
• Pile 4: A face-down pile in front of the BLACK pile containing
the same number of face-down cards the volunteer chose while
under the mind meld.

Remind everyone that your mind meld influenced the volunteer’s choice
of random cards and you can prove it. Even though the pack was shuffled
and the volunteer chose cards at random, you used the mind meld to
 The Red–Black Mind Meld: Abstraction, Algebra and Proof 257

Figure 38.1. The four resulting piles: Piles 1 and 2 face up with each card picked
by the magician to be of the right colour; Piles 3 and 4 face-down cards picked by
volunteer.

make sure they put exactly the same number of red cards in front of
your red pile as they put black cards in front of your black pile.
Ask the volunteer to first take their pile of face-down cards in front
of your RED pile and count aloud the number of RED cards in it onto
the table. Then ask them to take their face-down cards in front of
your BLACK pile, and similarly count aloud the number of BLACK
cards.
Through the mind meld, you caused the volunteer to select the
same number of RED cards in their RED-influenced pile as BLACK
cards in their BLACK-influenced pile. One card out and it wouldn’t have
worked!
258 Conjuring with Computation

The Magical Algorithm

TO DO the Red–Black Mind Meld:
1. DO an overhand shuffle of a full pack of 52 cards.
2. Ask for a volunteer to take part in a mind meld.
3. Deal the cards into two equal piles and hand one face down to
the volunteer, keeping the other.
4. Turn your pile face up.
5. Say you are beginning a mind meld from you to the volunteer.
You will control their actions even as they think they are making
their own decisions.
6. DO THE FOLLOWING UNTIL there are no cards left.
a. Select, at random, either some BLACK cards or some RED
cards from your pile.
b. Place them in one of two face-up piles in front of you, based
on the colour, saying how many cards you have chosen.
c. The volunteer selects the same number of cards without
looking at the actual cards, taking them from anywhere in
their pile.
d. They place their chosen cards in a pile in front of the pile
where you placed yours.
7. Remind everyone that your mind meld influenced the volunteer’s
choice of random cards and you can prove it. The pack was
shuffled, and the volunteer chose cards at random, but your mind
meld meant they put the same number of red cards in front of
your red pile as they put black cards in front of your black pile.
8. The volunteer counts the number of RED cards in the pile in
front of your red pile.
9. The volunteer counts the number of BLACK cards in the pile in
front of your black pile.
10. You note that as predicted they are the same number.

How It Works
Of course, it’s not mind melding. It’s just computation and mathematics.
The trick is self-working. As usual, if you follow the steps, it is bound
 The Red–Black Mind Meld: Abstraction, Algebra and Proof 259

to work because of the mathematics behind the steps the volunteer

performed. We can even prove it . . .

Thinking Computationally
So how can we be sure that this trick always works? Rather than doing
it lots of times (testing) or just using informal but logical reasoning, we
are going to use algebra.
The first step is to create a mathematical model of the trick.
By that we just mean describe it mathematically. Doing so involves
abstraction: focussing only on the information about it that matters
while ignoring anything that doesn’t. A mathematical model is just an
abstraction of a part of the world we are interested in. This trick is about
the numbers of red and black cards. The actual values and suits of the
cards do not matter at all. We can ignore that (abstract away those
details) and just focus on the count of each colour that ends up in each
pile.
We, of course, have no idea how many reds or blacks ended up in any
of the piles as the pack was shuffled. We can’t actually put numbers to
anything, apart from the fact that there are no black cards in the face-up
red pile and no red cards in the face-up black pile at the end. Instead,
we just give a name to the unknown numbers of red and black cards in
each pile. Then we refer to those numbers by name and do not worry at
all about the actual values. Those values, the number of cards in each
pile, will be different every time we do the trick. A mathematician calls
these names, variables, though the word here means something slightly
different to the kind of variables programmers talk about, it is a similar
idea of giving values names.
First, we give names to the number of red and black cards in each of
the four piles. First, consider Pile 1 (your face-up RED pile). We will call
the number of red cards there, RED1, and the number of blacks there,
BLACK1. The number of red and black cards in Pile 2 (your face-up
BLACK pile) will be called RED2 and BLACK2, respectively. Similarly,
we will say that the face-down Pile 3 contains RED3 reds and BLACK3
blacks. The face-down Pile 4 contains RED4 reds and BLACK4 blacks
(see Figure 38.2).
260 Conjuring with Computation

Figure 38.2. The four resulting piles with names given to the number of red cards
and number of black cards in each.

Logical thinking: The facts

The next task is to use logical reasoning to work out what facts we
actually know and turn them into mathematical equations about the
trick.
The first thing we know is that there are no black cards in Pile 1 as
you only put red ones there. There are also no red cards in Pile 2 as you
only put black cards there. That gives us two facts.
Fact 1: BLACK1 = 0
Fact 2: RED2 = 0
Now, we also know that a full pack of 52 cards was used. It contains
26 RED cards and 26 BLACK cards. If you add up the red cards in the
four piles, it will come to 26. We can write that as an equation using
the names RED1, RED2, RED3 and RED4 for the different sets of red
 The Red–Black Mind Meld: Abstraction, Algebra and Proof 261

cards in each pile as in Figure 38.2. We can do the same thing for the
black cards. We have to use the names here because we don’t know the
actual numbers.

Fact 3: RED1 + RED2 + RED3 + RED4 = 26

Fact 4: BLACK1 + BLACK2 + BLACK3 + BLACK4 = 26

We can, of course, make use of our first two facts, which do tell us two
of the numbers, to simplify Facts 3 and 4. Facts 1 and 2 say we can
replace RED2 with 0 and we can replace BLACK1 with 0 as they are
the same.
This gives new facts:

Fact 5: RED1 + RED3 + RED4 = 26

Fact 6: BLACK2 + BLACK3 + BLACK4 = 26

We also know something else about the numbers. Whenever we put

some number of cards in the red pile, we put exactly the same number
of cards in the face-down pile in front of it. That means that the two
piles always have exactly the same number of cards, though we have no
idea what that number is and it is different each time. We can again
turn this into a fact based on the names we have given each number
though. The number of cards in the RED Pile 1 (RED1) is the same as
the number of face-down cards placed in front of it in Pile 3. Pile 3 is
made up of RED3 red cards and BLACK3 black cards. That means that
together RED3 and BLACK3 must add up to RED1.

Fact 7: RED1 = RED3 + BLACK3

By a similar argument, we get a fact about the cards in Piles 2 and 4.

Fact 8: BLACK2 = RED4 + BLACK4

Logical thinking: Some rigorous reasoning

The last four facts sum up everything important we know about the
cards. They are a mathematical model of the trick.
262 Conjuring with Computation

Just as we did with the first two facts substituting BLACK1 for 0 and
RED2 for 0, we can now do a similar bit of logical reasoning substituting
for the things we know about RED1 and BLACK2 from Facts 7 and 8.
We replace RED1 in Fact 5 with RED3 + BLACK3 as they are
the same value. In the same way, we replace BLACK2 with RED4 +
BLACK4 in Fact 6. Our new versions of those facts are:

Fact 9: (RED3 + BLACK3) + RED3 + RED4 = 26

Fact 10: (RED4 + BLACK4) + BLACK3 + BLACK4 = 26

Now we have two different things that we know equal 26 so the left-hand
sides of both those equal signs equal each other. Combining Facts 9 and
10, we get

Fact 11: RED3 + BLACK3 + RED3 + RED4

= RED4 + BLACK4 + BLACK3 + BLACK4

Things seem to just be getting more and more complicated, but we

can now start to simplify things. There is a RED4 on both sides of this
equation so they cancel. Put it another way if we subtract RED4 from
both sides (so keeping both sides equal still), we get

Fact 12: RED3 + BLACK3 + RED3

= BLACK4 + BLACK3 + BLACK4

Similarly, BLACK3 appears on both sides so can be cancelled.

Fact 13: RED3 + RED3 = BLACK4 + BLACK4

We have two lots of RED3 on one side and two lots of BLACK4 on the
other. That can be written

Fact 14: 2 × RED3 = 2 × BLACK4

Then finally, dividing both sides by two, we get a really simple fact
that is guaranteed to be true as it followed mathematically from our
original facts.
 The Red–Black Mind Meld: Abstraction, Algebra and Proof 263

Fact 15: RED3 = BLACK4

This tells us that the number RED3 will always be the same as the
number BLACK4 at the end of the trick.

Back to the real world

Now that is all well and good, but what does it mean? We have to return
from our mathematical world of variables to the real world of piles of
cards. What do those variables actually represent? RED3 and BLACK4
were actually just names that stood for numbers of cards of particular
colours in the face-down piles. RED3 is the number of red cards in the
red-influenced pile, Pile 3. BLACK4 is the number of face-down black
cards in the black-influenced pile, Pile 4.
The algebra has shown that the number of red cards in Pile 3 is
guaranteed to be the same as the number of black cards in Pile 4 at the
end of the trick. That is exactly what we predict in the trick . . . because
it is guaranteed to be true by the algorithm we followed.

Verifying programs
If we can prove that a magic trick always works, then we can prove
that a program does too in a similar way. We use abstraction to
create a mathematical model of what the program does. We then
use algebra to prove properties about the model. Those properties
tell us facts about what the program does in the real world. This is
called program verification. For example, we might prove that if the
program controlling a medical device was set to deliver a fixed number
of milligrams of a drug over a fixed number of hours, then that is what
it always does, whatever the actual numbers involved.
When lives depend on a program always working, it is no good
that we tested the program and decided we were 99% sure it won’t
kill anyone, especially if it is used by millions. Proofs can tell us that a
program always works (as well as showing that our tricks do).
264 Conjuring with Computation

PAUSE: Shafi Goldwasser

Shafi Goldwasser’s work is about using maths and proof to make
computers more secure and private. She studies problems that are
so hard that computers would take billions and billions of years to
solve them. These are the kinds of problems that the encryption
used to secure computer systems is built around. Along with Silvio
Micali and Charles Rackoff, she came up with the idea of zero-
knowledge proofs. These are a kind of proof that at first sight seems
impossible: something that would take actual magic to achieve. The
idea is for one person (or computer) to prove that they know some
secret information while actually keeping that information a secret.
They do not give up the secret even to the person (or computer)
they prove they know it too. It is like having a password that lets
you into a computer, but without you actually having to type the
password. You prove you know it without ever actually revealing it
even to the computer doing the checking.
 Part 10

More on Computational Thinking

Computational Thinking is a collection of separate skills that we

have seen throughout the book including decomposition, abstraction,
generalisation, choosing good representations and logical thinking. In
this section, we look at some more advanced computational thinking
approaches based on them.
This page intentionally left blank
 Chapter 39

The Doomsday Clock

Generalisation

Conjuring
The Doomsday Clock predicts the most important hour in your
future. Even with a free choice to start, its hour hand ends on a
predicted hour.

Computation
By abstracting away detail until we get to the very core of a trick,
or algorithm, so its most general version, we can then build it back
up into new tricks or algorithms.

The Trick
Before you start this trick, you need to create a big clock face with
numbers from 1 to 12. Make a single movable hour hand for the clock,
drawing a fixed minute hand pointing to 12 (see Figure 39.1). You also
need to draw a second smaller version of the clock on a piece of paper
with only the number 1 at the 1 o’clock position, and with a fixed hour
hand pointing at the 1 (see Figure 39.2). This is your prediction. Place

267
268 Conjuring with Computation

Figure 39.1. The Doomsday Clock. The minute hand is fixed. The hour hand moves.

Figure 39.2. Your prediction for the time the Doomsday Clock will stop.

it in an envelope, fastened to the back of the clock. Place the hand of

the big clock on 12, and you are ready to start the trick.
Explain that you have made a Doomsday Clock: a clock that can
predict the hour that will be most important to someone in the future.
 The Doomsday Clock: Generalisation 269

It may be the hour they will get married, their first baby is born, or
perhaps the hour they will die. Who knows? What is for sure it will be
an hour in the future that will be really important to that person’s life.
First, ask for a volunteer. Explain that you will use the clock to find
the hour in their future that will be most important to them personally.
It will take only a few turns of the hour hand until the hour hand passes
12 once more. Ask them to choose a number from 1 to 12. Write it
down as a word (so if they choose 2, write T-W-O). Have them then
move the hand of the clock one place round the clock for each letter in
their chosen word. If they choose TWO, for example, then they move
T-W-O so three places on. Tell them they will be using the words of
numbers like this to jump the hand forwards three times. They are now
at a new position on the clock and at a new number. This gives a new
word to use to count further. They do the same again with that word.
They then do it again a final time at the new hour they land on, and
then stop the clock at that point.
At whatever place they end up, announce that their personal clock
stopped at that time. Point out it was their free choice of the original
number. Now, explain that, to show the power of the clock you already
used it to do the prediction for them. Turn the clock over revealing the
envelope with your prediction. Ask them to open the envelope and read
the prediction of their doomsday hour. Amazingly, it is the same as on
their clock. That is the power of the Doomsday Clock!

The Magical Algorithm

TO PREPARE The Doomsday Clock:
1. Create a clock face with numbers 1 to 12 for the hours and a
single movable hour hand and fixed minute hand pointing to 12.
2. Set the clock’s hour hand to 12.
3. Create a clock face with a fixed hour hand pointing to the hour 1
as a prediction.
4. Place the prediction in an envelope fastened to the back of the
clock face.
270 Conjuring with Computation

TO DO the Doomsday Clock:

1. A volunteer chooses a number between 1 and 12: one of the hours
of the day.
2. Write it down as a word.
3. DO THE FOLLOWING THREE TIMES.
a. The volunteer moves the hand on one hour, clockwise round
the clock, for each letter in the current word.
b. Write down as a word the hour that the clock hand ends on.
4. Announce the final hour that the clock ends on.
5. The volunteer opens the envelope to reveal your prediction.

How It Works
This trick is based on the same property to do with the randomness of
the length of words from the Wizardly Book Magic trick of Chapter 6.
It means in this case that in three steps all 12 numbers you can choose
take you to the same place. The length of words does not actually vary
very much. The words in the clock all have lengths between 3 and 6
letters (see Figure 39.3). As it is the length of the word we use rather
than the word itself, that means the original 12 numbers to freely choose
from actually only give four possibilities: lengths of 3, 4, 5 or 6.
Since we always start at 12, after counting out the first word, there
are only four possible places we could be at after the first step: the hours
3 to 6 (see Figure 39.4). Now the length of THREE is 5, but the length

Figure 39.3. The lengths of words corresponding to numbers on the clock face.
 The Doomsday Clock: Generalisation 271

Figure 39.4. The possible ways the clock might change.

of FOUR one place after it is only 4, so they will both count you to the
same next place: hour 8. The lengths of FIVE and SIX have the same
pattern, and both take you to the position of hour 9. There are now,
after 2 steps, only two places you could be. Those positions, 8 and 9,
have the same pattern of length, as the previous ones, with the second
being one less letter in length than the first. They therefore both jump to
the same place: 1. So after only three steps, all options have surprisingly
led to the same place. That is your prediction.
Rather than telling the volunteer to stop on the third step, you could
alternatively say they will stop when they have passed 12 once more as
this is equivalent.
272 Conjuring with Computation

Thinking Computationally
Once you understand the basic principle of a trick, you can reuse that
principle as the basis of different tricks. This trick uses exactly the same
principle as Wizardly Book Magic (Chapter 6). Rather than using words
in a book, here we are just using the words for the numbers on a clock.
If you have invented a trick like Wizardly Book Magic, then strip it
down to its core principle (abstracting away the details), and you have a
generalised version of the algorithm. You can then build a new trick by
providing a new context. Figure 39.5 gives a generalised version of our
two variations of this trick. It is just a description of the mathematical
principle behind the trick as an algorithm.
Programmers use generalisation like this a lot. Rather than writing
code or creating an algorithm just for one specific purpose, once we have
solved a problem in a new way, we generalise it so it can be used in lots
of situations.
If we have written code to search for names in an alphabetically
ordered list of names (e.g., in an online telephone directory), we might
generalise it so that it can also be used to search for other text that
is ordered alphabetically. We might generalise it further, modifying the
code so that it can be used to search for other data ordered in some way
(e.g., numbers in numeric order). The increasingly generalised versions
have a wider use and are more abstract versions of the original concrete
code. They all are based on the same underlying mathematical principle
just as in our trick.

1. A volunteer chooses a starting point in some sequence of words.

2. DO THE FOLLOWING UNTIL you get beyond a preset point, after the point that
all paths meet:
a. Count the letters in the word.
b. Move on that many words.
3. Announce the final word that the volunteer lands on.
4. Reveal a prediction of the word they ended up on.

Figure 39.5. A generalised version of the Doomsday Clock and Wizardly Book
Magic.
 The Doomsday Clock: Generalisation 273

PAUSE: John von Neumann

To computer scientists, John von Neumann (1903–1957), one of
the founders of computer science, is most famous for his paper
describing the basic structure of a computer. That structure is
now known as the von Neumann architecture, and most modern
computers follow the design. It describes a computer as split into
a processing unit to do calculation and a control unit that controls
which instructions are executed. Data and programs are stored in
the same computer memory. It was actually based on the ideas of
J. Presper Eckert and John Mauchly for their ENIAC computer.
Von Neumann’s contributions to the subject were very wide and
included, with Herman Goldstine, the suggestion of using assertions
to indicate properties that hold at a point in a program. He was a
Physicist and Mathematician working on the Manhattan Project to
develop the first nuclear bomb. He also came up with the idea of
mutually assured destruction, arguing nuclear weapons would never
be used in a war if both sides knew everyone would die if they were.
This page intentionally left blank
 Chapter 40

Free My Three
Divide and Conquer

Conjuring
Three cards of a kind chosen by a volunteer are lost in a pack.
After repeatedly discarding cards with no sign of the chosen ones,
they are eventually revealed as the last cards left.

Computation
Decomposition involves breaking a problem into smaller problems
that are easier to solve. Divide and conquer is a powerful version
of this where these smaller problems are the same problem as the
original (just smaller). Those smaller problems are solved in the
same way. This gives a way to create very fast algorithms.

The Trick
For this, you need a full, shuffled deck of 52 cards. Ask a volunteer to
find and remove three cards that are three of a kind, such as three of
the queens. Have them show the audience, but not you. Everyone must
remember those three cards.

275
276 Conjuring with Computation

Take the pack back, without those cards, give them a shuffle and
then casually make three piles. The piles must contain 10, 15 and
15 cards. It’s fairly easy to push off blocks of three cards at a time.
The first pile, with 10 cards is three blocks of three with one extra. The
15-card piles are simply five blocks of three. You will be left with nine
cards in your hand (but your audience won’t know that as it looks like
you have just built three random piles. Pop these left-over nine cards to
the side for the moment.
The three selected cards are put onto the three “random” piles and
“lost”, but in a special way (see Figure 40.1). The first card goes onto
the 10 pile. Next, ask a spectator to cut off a chunk of cards from
the adjacent pile, the 15 pile, and pop them on top of the chosen card
burying it.
The next chosen card goes on top of the pile they have just cut the
cards off. Again, once they place the chosen card on top, they bury it
by placing a chunk of cards from the remaining 15-card pile onto it.

Figure 40.1. How the three chosen cards are lost in the pack.
 Free My Three: Divide and Conquer 277

The final chosen card is placed onto that last pile, and you then
drop the nine-card pile on top.
Next, assemble the deck, starting with that last pile you put the nine
cards onto. Put this pile on top of the pile next to it and then put this
bigger pile on top of the remaining pile. Square the cards.
Now, remind the audience what has happened: you made three
random piles, and each of the chosen cards was put into a pile and
buried under a random stack of cards from another pile, so no one has
any idea where any of the chosen cards are . . . But you are now going
to try to find them and free them from the pack.
You deal the cards into two piles. Start by dealing two piles of two
cards each face down, by alternating left, right, left and right. Then
stop and say that actually you should make the trick more dangerous by
dealing with one of the piles face up. Collect the four face-down cards
from the table and put them at the bottom of the pack. What you have
just done in this tester deal is shifted four cards from the top to the
bottom of the pack, but also built up the tension.
Start again by dealing a card face up. Then, deal the next face down,
telling your audience that the trick ends if any of the three selected cards
goes in the face-up pile. Deal through the whole deck like this. No chosen
cards will be seen.
Put the face-up pile aside, take the face-down pile, and deal it again:
first card face up, next face down and so on. Again no chosen card will
be seen. Repeat this procedure, dealing the first card face up each time,
until the face-down pile contains only three cards. Ask for the names of
the three chosen cards. On turning the last three cards over, the three
selections are found!

The Magical Algorithm

TO DO Free My Three:
1. DO set up the deck.
2. DO the teaser deal.
3. DO the deal.
278 Conjuring with Computation

TO DO set up the deck:

1. A volunteer finds and removes three matching cards from a
shuffled pack.
2. They show them to the audience.
3. DO an overhand shuffle of the remaining cards.
4. Place those remaining cards into piles of 9 cards, 15 cards and
15 cards, face down.
5. Set the remaining cards to one side.
6. The volunteer places one of their three cards onto the first pile
of nine cards.
7. The volunteer cuts the top of the next pile and places those cards
on top of the chosen card.
8. The volunteer places the second of the three cards on top of the
middle pile.
9. The volunteer cuts the top of the last pile and places those cards
on top of that second chosen card.
10. The volunteer places their final card on top of the last pile.
11. You place the final pile of cards you placed to one side on top of
that card.
12. Place the third pile on top of the second to create a larger second
pile.
13. Place this larger second pile on top of the first to create a new
full deck of cards.
TO DO the teaser deal:
1. Deal out four cards into two piles face down.
2. Say “Actually let’s make this more dangerous. Let’s have one pile
face up.”
3. Collect up the four cards dealt and place them at the bottom of
the deck.
TO DO the deal:
1. DO THE FOLLOWING 4 TIMES.
a. Pick up the remaining cards.
b. DO THE FOLLOWING UNTIL there are no cards left.
i. Deal a card face up.
ii. Deal a card face down.
c. Discard the face-up cards.
 Free My Three: Divide and Conquer 279

2. Turn over the three remaining face-down cards to reveal the

originally chosen cards.

How It Works
You started with four piles of 10, 15, 15 and 9. The first card went into
position 11. Then some of the next 15 cards were added on top. This
appeared to add some randomness, but actually did nothing important
as the remaining cards were later just put back on top of them. That
meant there were actually exactly 15 cards separating the first secret
card from the next secret card, as it was added to the top of the second
pile. This was done again with the next pile meaning another 15 cards
separated the second and third secret cards and finally, another nine
were placed on top.
At this point, the secret cards, apparently buried randomly, are at
known positions from the top: positions 10, 26 and 42. However, in the
false start, we actually move the top four cards from top to bottom.
This shifts everything up four positions. So when you start the actual
deal, the secret cards are at positions 6, 22 and 38 from the top.
That whole palaver was actually a way to take a shuffled pack of
cards, but leave the three secret cards at those fixed positions while
making the audience think they had been buried randomly.
Why those positions? Well, each deal actually filters out every second
card. However, it also reverses the order. It turns out that if you do that
repeatedly you will end up with exactly the cards that started in positions
6, 22 and 38. We can see how by working through what happens . . .
We start with cards as the following where we just note their initial
positions:
1, 2, 3, 4, . . . , 50, 51, 52.
In the first deal, we start by placing a face-up card, and as we ultimately
throw away the face-up cards, we throw away every second card starting
with card 1. That means we lose all the odd-positioned cards. We also
reverse the order as we do so leaving a new pile:
52, 50, 48, 46, . . . , 8, 6, 4, 2.
280 Conjuring with Computation

We do this again, so again take out the current first, third, fifth, etc.
cards. This leaves, once reversed, the cards originally in positions:
2, 6, 10, 14, 18, 22, 26, 30, 34, 38, 42, 46, 50.
Another round of the deal takes out every second card and reverses
again, leaving:
46, 38, 30, 22, 14, 6.
Finally, taking out every second card one last time leaves us with the
cards that were originally in positions:
6, 22 and 38.
Those are exactly the positions we made sure the secret cards were at
before the deal started, ensuring the trick was guaranteed to work.

Thinking Computationally
The core of this trick involves the final series of deals, where we
repeatedly strip out every second card. This process guarantees we end
up with the given three cards. If we went a step further, we would
guarantee to have a single card. This magic algorithm is linked to an
important decomposition principle called divide and conquer. It is used
to create fast algorithms.
You can think about the trick as being an algorithm to find those
selected cards. Put that another way, it is about how to discard all the
cards aside from the selected ones.
We do that by discarding half the cards on each pass. It only takes
four passes to get down to the three secret cards. There is another
interesting feature of the way this is done. At the start, you are holding
a pack of cards and want to be left with the secret cards. Having done
one deal and thrown away half the cards, you are in the same situation:
you are holding a pack of cards and want to be left with the secret cards,
so you do exactly the same thing again. That leaves you in the same
position . . . so you do the same thing again. Only the number of cards
in the pack changes each time: the pack so the problem gets smaller.
This is a special way of decomposing a problem. It has been
decomposed into a smaller, simpler version of the same problem, not just
 Free My Three: Divide and Conquer 281

created a smaller but different problem. Because it is the same problem,

we can decompose it again in exactly the same way, and keep doing that
until it is a trivially simple version of that original problem. In the trick,
that trivial version is we want to find three cards and are holding a pack
of three cards (trivial!). This process is called recursion. It is actually,
computationally, just a version of repetition, as it involves repeating
the same computation over and over. This is the basis of recursive
problem solving: finding a way to solve a problem by decomposing it
into a smaller version of itself.

Divide and conquer problem solving

Recursive problem solving leads to an even more important decompo-
sition idea called divide and conquer problem solving. Rather than
finding a slightly smaller version of the problem, this involves finding a
way to break a problem into identical problems that are a fraction of
the original size. This might involve turning it into one smaller problem
that is a fraction of the original size, or two similar problems that are
half the size, or three versions that are a third of the size, and so on.
As the problem is the same, just massively smaller, we can keep solving
the new problems generated in the same way.
Divide and conquer problem solving is a very general way of devising
very fast algorithms.
What is happening here is similar. On each pass, we get rid of not
just a few cards but half the cards. Each pass leaves us with a single
version of the equivalent problem that is half the size by discarding half
the cards. We very quickly (in terms of the number of deals of the
remaining deck at least) are down to the three cards we are looking for.
This trick doesn’t in itself really show off the power of divide and
conquer, because we actually spend a lot of time splitting off each card.
A variation of the algorithm is in fact the basis of a true divide and
conquer algorithm though, where every second card is removed in one go.
It was used by early computers to process punch cards: an early way of
storing data.
To see the real power of divide and conquer, take a pack of shuffled
cards and put them back in order. How? There are lots of different ways
282 Conjuring with Computation

(i.e., algorithms) to do it. One way, e.g., is to first scan down the pack
looking for the first card (the Ace of Hearts), then scan down it looking
for the next (the two of Hearts), and so on. This is very slow. You scan
down the remaining pack, on average looking halfway for each card, 51
times. You will likely check individual cards over 1,000 times.
Another way is to start by running through the pack, putting the
cards into piles (like in the trick) but one pile for each suit of Hearts,
Spades, Clubs and Diamonds. That takes a single scan. We are now left
with four versions of the same problem each a quarter of the size. We
have the start of a divide and conquer approach! To make it work though,
we must be able to do the same again. We must run through each pile
in turn and split them into four piles too, each leaving a problem of
sorting them. That’s easy: a pile for Ace to 3, a pile for 4 to 7, a pile
for 8 to 10, and a pile for the court cards. Each has three or four cards.
Now taking one of those piles, we do the same again, with three or four
piles, smallest to largest again, but this time we end up with piles of a
single card. That is trivial to sort! A pile of one card is already sorted.
Now we put those four cards on top of each other in the order of the
piles and now have those four cards sorted. We move back to the next
pile of four and sort it in the same way. It can be placed on top of the
first, and so on. Eventually, we have sorted all the Hearts. So we move
on to the Spades and do exactly the same, and so on for each suit.
We sorted each pile using the same general divide and conquer
method. Split it into four piles and sort them the same way. Once sorted,
pile them back up to give a bigger sorted pile. Get a pack of cards and
try it.
It probably still sounded slow, but it is actually much faster because
rather than working through all the cards lots of times, we have broken
them into smaller and smaller piles. The key is that we get down to a
single card in only three steps. This essentially means that instead of
scanning the whole pack 51 times, we only do so three times: once to
split them into suits, once to split them into groups of four, and once to
split them into single cards. It is therefore massively faster and involves
scanning cards only a few hundred times instead of over a thousand.
 Free My Three: Divide and Conquer 283

Radix sorting
Imagine, now, you have not 52 cards but 5,000 cards (or perhaps, to
face a problem I have repeatedly faced in the past, they are exam scripts
that need sorting). Each has a four-digit number on it, like 1,234, and
you need them to be sorted in order of that number. We can use exactly
the same idea but use the different digits to organise each scan. On the
first scan, put cards into 10 piles based on the highest (thousands) digit.
Next, we take one of those piles and split them based on the hundreds
digit. Then do the same on each pile for the tens digit and finally for the
units. At that point, we have piles of one script, with the piles in order
so we just pile them on top of each other. We then move on to the next
pile and so on. This is called radix sorting. It is just one of many divide
and conquer solutions to the problem of sorting data into order.

Interlude: Alex Elmsley

“Free My Three” is a variation of a famous effect invented by magi-
cian Alex Elmsley (1929–2006). He was a graduate of Cambridge
University where he studied Mathematics and Physics, before going
on to work as a patent agent and then computer programmer. There
are very many variations built on the underlying mechanism: see
Chapter 49 for another variation of this effect. He also invented
a series of other card tricks and wrote about the mathematics
underpinning shuffling cards.
This page intentionally left blank
 Chapter 41

The Double-Destination Deception

Data Representation — Graphs
and Cycles

Conjuring
An audience member constructs a trip-of-a-lifetime flight itinerary
that crisscrosses the world. They plan it by making a line of tickets.
Despite their freedom to choose any itinerary, once completed,
you reveal you bought a ticket in advance. Remarkably, it takes
them back to their chosen start destination from the place they
ended up.

Computation
Graphs are an important data representation used to store and
manipulate data when the links between different pieces of data
matter. They are also the basis of logical reasoning about whether
it is possible to write an algorithm to do certain tasks.

The Trick
Before doing the trick, you need to create a series of tickets. There are
two kinds of tickets: flights and hotels. Flight tickets show the start and

285
286 Conjuring with Computation

Figure 41.1. Example hotel and flight tickets.

end destinations of flights; hotels just show a city name (see Figure 41.1).
The full set of flights and destinations is:

Hotel tickets:
London, Singapore, New York, Delhi, Sydney, St Petersburg, Beijing
Flight tickets:
London–Singapore New York–London
New York–Singapore Delhi–London
Delhi–Singapore Delhi–Singapore
Delhi–New York Sydney–London
Sydney–London Sydney–Singapore
Sydney–New York Sydney–New York
Sydney–Delhi St Petersburg–London
St Petersburg–London St Petersburg–Singapore
St Petersburg–New York St Petersburg–Delhi
St Petersburg–Sydney St Petersburg–London
Beijing–Singapore Beijing–New York
Beijing–Delhi Beijing–Sydney
Beijing–St Petersburg

As long as you have seven cities with a flight ticket for every pair and a
hotel for every city, you can actually use any city. Place one of the flight
tickets in an envelope, and you are ready to start the trick.
Give the envelope with the selected ticket to a member of the
audience to look after saying that it holds a special prize ticket. Ask
another volunteer to use the remaining tickets (flights and hotels) to
plan an itinerary crisscrossing the world. They should place them in a line
to show their planned route. Explain that it doesn’t matter where they
start and finish. Flights can be used to fly in either direction. Each ticket
 The Double-Destination Deception 287

must be used exactly once. The volunteer should match up start and
end cities of adjacent tickets as they must also always arrive and depart
from the same destination: no overland treks from one city to another!
Hotel tickets can be used at any time when they are in that city, though
they can only stay in a hotel in each city once. At other times, they are
just in transit elsewhere.
Once they have created their itinerary, point out they had a
completely free choice about the order of the tickets including where
they started and ended. Have them write down their start and end
destinations so everyone can see. Point out that it is a bit of a shame to
do a round-the-world trip but not end up where you started. Luckily, you
thought of this for them. Have the person holding the envelope reveal
your pre-bought ticket. Amazingly, it is a ticket between the volunteer’s
start and end locations. It will complete their personal round-the-world
trip.

The Magical Algorithm

TO PREPARE the Double-Destination Deception:
1. Create a full set of tickets for seven cities consisting of seven hotel
tickets and 25 flight tickets, one between each pair of cities.
2. Place one flight ticket in an envelope marked “Special Prize
Ticket”.

TO DO the Double-Destination Deception:

1. Give the envelope with the prize ticket to a volunteer.
2. A second volunteer creates a line of the tickets to be their itinerary
crisscrossing the world so that the start and end destinations of
each pair of adjacent tickets match.
3. The volunteer writes their final start and end destinations for the
whole trip where everyone can see.
4. You point out that “You had a free choice of the itinerary, and it
would be a shame not to end up back where you started. Luckily,
I thought of that.”
5. The first volunteer opens the envelope containing the special
ticket, to reveal that it is exactly the ticket they need to take
them back to their start point.
288 Conjuring with Computation

How It Works
An alternative way to do this trick is using a set of dominos. The trick is
based on a little-known mathematical fact about a full set of dominos.
You can place all the domino pieces in a circle, matching the numbers
on the tiles, and this line will in fact join up making a complete cycle
with all adjacent dominos matching. The actual order of the pieces may
vary, but if you have a full set, you can always find a chain that will
join with itself. The tickets in the trick correspond to dominos. A ticket
linking each pair of destinations (the flights) is like a normal domino with
different numbers. A hotel ticket corresponds to a “double” domino with
the same number twice.
Knowing this, it’s easy to see how the trick works. You have removed
one flight ticket so have removed one link in the cycle. The volunteer
won’t notice one piece is missing, but when they are challenged to put
all the tickets in a single chain, the end of their chain will be missing that
link that joins it back to the start. When the ticket is revealed, you are
just completing the cycle. The destinations will match your prediction.
Now you know the pattern that makes it work, you can of course
generalise it in your own way and so transfer this basic idea for a
prediction trick by using your own cycle of data. For example, rather
than destinations or the two numbers on a domino, one of the attributes
of your piece could be a colour and the other a letter say. Or a piece
could have just text on it like a singer’s name with one of their songs for
a music top 40-type presentation. Just make sure that at the beginning,
you write down the full cycle on paper before you transfer the numbers,
letters, songs or whatever, to individual pieces.

Thinking Computationally
We’ve said that this is a property of dominos, but why does that property
hold? To a computer scientist, this is a graph problem. Graphs are a
data representation, used as an abstraction of real-life things. They make
that original thing (whether dominos or flights) easier to compute with,
model or reason about. Road networks, communication networks and
social networks (and dominos) are all things that can be represented by
a graph. We have actually used graph data representations to explain
 The Double-Destination Deception 289

tricks in this book already (see, for example, Figure 39.4 in Chapter 39,
the Doomsday Clock). A graph consists of points or nodes (think of
these as places) and lines or edges (links between those places). So our
cities can be represented by the nodes of a graph, and the tickets or
flights linking them can be represented by edges. Nodes don’t have to
be actual places though. Each number in a domino set (from 0 to 6)
can be represented by a node. Each domino in a set is then represented
by an edge linking two different numbers or a number to itself.
Now let’s start thinking about why our trick works by looking at a
simpler version with only three cities: London, Singapore and New York.
We then have six tickets as follows:
Hotels:
London, Singapore, New York
Flights:
London–Singapore
London–New York
Singapore–New York

Drawing these as a graph, we get a clearer picture of what is going

on (Figure 41.2). Hotels are just drawn as edges from a place to itself.

Figure 41.2. The graph representing the tickets when there are three cities.
290 Conjuring with Computation

Figure 41.3. The graph representing the tickets when there are four cities.

Each node has two edges leading into it and two leading out. There is
an obvious cycle. Removing any flight would break the cycle, and the
ends of the chain would correspond to that flight. The trick would work
(though not be so magical).
However, if we now add an extra city with a hotel and a flight from
it to each other city, things go wrong (see Figure 41.3). The tickets are
now as follows:

Hotels:
London, Singapore, New York, Delhi
Flights:
London–Singapore London–New York Singapore–New York
Delhi–London Delhi–Singapore Delhi–New York

It turns out that it is impossible to create a cycle with four cities

(try!). Next, add another city and try again before reading on.
As we keep adding a new city, so its hotel and the flights from it to
everywhere else, a pattern seems to emerge. For some, you can create
a cycle (3, 5 and 7 cities). For some, you can’t (4, 6 and 8 cities).
 The Double-Destination Deception 291

The Bridges of Königsberg

At this point, we take a detour and go to the very birth of the idea of
a graph. It is 1736 and Leonhard Euler has solved a puzzle about the
Prussian town of Königsberg. The river Pregel runs through the middle
of the town. In the river are islands connected by bridges to either bank
(see Figure 41.4). The puzzle is to find a way for a tourist to complete
a walking tour of the town crossing each bridge exactly once. We will
consider a variation where you must also end back where you started.
Euler realised that he could simplify the problem (always a good
idea) by drawing the banks and each island as circles and the bridges
between them as lines (Figure 41.5). He had invented the abstraction
of the graph. In computer science terms, this is a data representation
or data structure, but a very general one. Simplifying things down to a
graph allowed Euler to see the answer to the problem. Our problem has
become a question of whether you can find a cycle in the graph that
passes along every edge (the bridges) exactly once returning to where
you started. He had generalised the problem to one about graphs.
Euler realised that there was an easy, and general, way to work out
whether paths like this were possible or not in any graph. Look at any
one node. For a cycle to pass through that node, every route (edge)
in must have a different route (a different edge) out. The edges of the
graph at each node must pair up. If there are an even number of edges

Figure 41.4. The Bridges of Königsberg map.

292 Conjuring with Computation

Figure 41.5. The Bridges of Königsberg as a graph.

Figure 41.6. A node can only be in the middle of a path (or part of a cycle) if there
are an even number of edges to it so that for every way in there is a way out. Here
the extra edge makes an odd number, so if followed into the node, there is no way
left to leave that node.

in the node, that is possible. If there are an odd number of edges in a

node, then it is impossible. In that impossible case, one edge will always
be left unused. It will be a way in with no way out to carry on the cycle
(see Figure 41.6). For a cycle visiting all nodes, this rule must apply to
all the nodes: all nodes must have an even number of edges.
In the Königsberg graph, all the nodes have an odd number of edges
(three or five edges) into them. Therefore, by Euler’s theorem, a cycle
is impossible. There is no algorithm that would take you on a tour that
crosses all bridges, visit all islands and back to the start. It is actually
 The Double-Destination Deception 293

impossible to do even without returning to the start as Euler proved

with similar reasoning.

Back to the magic

We can apply the same logical reasoning to any graph. That includes
the graph of our magic trick. If we look at the graph for a version of the
trick with three cities, each city is connected to the two other cities and
itself. Ignoring for a moment the hotel edges that come and go from the
same node, there is an edge in and an edge out of every node: an even
number. There is an obvious cycle. Adding in the hotel links doesn’t
change that as even though each hotel link is a single edge it counts as
a pair: an edge going in and out.
When we add a city to make four cities, we have added a new edge
into each node linking it to the new city. Each node now has five edges
instead of four (just as in Figure 41.3). That is an odd number, so by
applying Euler’s theorem, we immediately know there is no cycle. The
real trick used seven cities, meaning eight edges at every node (six to
link to each other city and a hotel that goes in and out). That is an
even number of edges, therefore there is a cycle.
Now, what happens when we remove one flight as in the trick? We
are removing one edge. The nodes of those two cities now have one
less edge. There are now two nodes with an odd number of edges, so
any chain of links will have to start and end at those nodes. We said
that the ticket removed by the magician should not be a hotel. We can
now see why. If we remove a hotel, we are removing two links from a
single node. It remains even and there will still be a full cycle, and the
volunteer could start and end it anywhere.

Dominos
We noted that the trick also works with dominos instead of tickets.
Follow similar reasoning and draw a graph for the possibilities with
dominos. Exactly the same graphs result, so all our above logical
reasoning works exactly the same for a domino version too, as long
as you use a full set of dominos with tiles with 0 to 6 spots. Try it for
yourself.
294 Conjuring with Computation

Interlude: Dynamo
Magician, Dynamo (real name Steven Frayne), baffled the audience
on his TV show with a variation of this trick. He grew up in a
deprived area of Bradford. His grandfather taught him his first trick,
a way to make himself look bigger to deter other children from
bullying him. He is one of the first-generation Internet magicians:
posting online videos of his street magic helped launch his career.
His stage name, after Faraday’s invention of a device that produces
electricity, came from an audience member who shouted out that
Frayne was like a dynamo during a performance he gave when he
was still a teenager. It was part of celebrations for the legendary
magician, Houdini’s centenary.
 Part 11

Cyber Security, Privacy and Society

Spies and hackers exploit both legitimate and covert channels in

computer systems: ways that information can leak from the system
without anyone knowing. Magicians similarly build tricks around such
channels: ways that information can leak from the magical system to
the magician without the audience realising. Magicians often pretend to
know amazing things about us, and some charlatans use tricks to con
people. Technology is taking away our privacy for real and often using
information gained in bad ways. Skills and technology should be used
for good, not harm, which is why professional ethics matter.
This page intentionally left blank
 Chapter 42

Sniff Out that Card

Steganography

Conjuring
A volunteer chooses a card and holds it to their chest while
memorising it. They then place it back in the middle of the pack
of cards while you look away. You work through the pack, smelling
the backs until you find the one that they picked, detected due to
the trace smell left behind.

Computation
Steganography is a way of hiding secret messages in apparently
innocuous things. What looks like a normal message or image
disguises the real thing being communicated.

The Trick
You shuffle a pack of cards thoroughly using a normal overhand shuffle
(Chapter 15) in front of the audience. You then fan the pack and ask
a volunteer to pick a card. They pull it out, look at it to be sure what
it is, and then hold it against their heart for a few moments while
they memorise it. You then hold out the pack for them to reinsert it

297
298 Conjuring with Computation

in a random place. You shuffle the pack again, and then holding the
pack with the faces to the audience, work through them a few at a time
sniffing the backs, searching for the smell of their card. Triumphantly,
you pull out a card, and it is theirs.

How It Works
You need a special pack of cards for this trick that has a picture or
pattern on the backs that is not symmetrical. Normal packs of cards
have symmetrical patterns, but it is fairly easy to buy a souvenir pack
of cards that either has a picture on or some other pattern that is
not symmetrical. Magic shops also have cards with patterns that look
symmetrical at a casual glance but actually are slightly different at the
corners.
Before the trick, you ensure all the cards are the same way around.
Then, when the person takes the card and is holding it to their heart, you
surreptitiously turn the pack around in your hands. When they reinsert
the card, it becomes the only card the other way around. As you go
through the cards supposedly smelling them, you are actually looking
for the one card that has been turned around (Figure 42.1). When
revealing the card, put the rest of the pack face up on the table in
a neat pile, the right way around and keep the back of their chosen card
facing you so no one can spot what is going on.

The Magical Algorithm

TO PREPARE Sniff Out that Card:
1. Take a pack of cards that contain a non-symmetrical image on
the back and turn them all the same way.
TO DO Sniff Out that Card:
1. DO an Overhand Shuffle of the pack of cards.
2. Fan the pack.
3. The volunteer chooses a card.
4. They look at their card, show it to the audience, remember it,
and hold it to their heart.
5. You turn the pack around.
 Sniff Out that Card: Steganography 299

Figure 42.1. (a) Cards with a non-symmetrical pattern look different when turned
around. (b) This allows a single card that is turned the other way to the rest to be
easily found.
300 Conjuring with Computation

6. They insert the card back somewhere in the middle of the pack.
7. DO an Overhand Shuffle of the pack of cards.
8. Hold the pack with the faces of the cards to the audience.
9. DO THE FOLLOWING UNTIL you come to a card that is turned
the wrong way around:
a. Look at the back of the card to see if it is turned the right
way around.
b. Sniff the back of the top card.
c. IF the top card is not the wrong way around.
THEN move it to the bottom.
10. Place the pack face up and neatly squared on the table apart
from the top card.
11. Reveal the top card as their chosen card.

Thinking Computationally
Steganography is a way of ensuring a message, and more generally
information, is secret by hiding it in some other innocuous thing. The
aim is that no one realises there is a secret message there at all. One
of the earliest versions from Ancient Greece involved a message being
tattooed on the shaved head of a messenger. They had to wait for their
hair to grow back before taking the message. Mary Queen of Scots also
used steganography. She famously plotted while under house arrest by
hiding messages in barrels of ale.
Information can be hidden in apparently innocuous messages in all
sorts of ways. Different patterns in an image could stand for different
letters, or a subtle difference in the way letters are written could stand
for 1s and 0s with a message then stored as binary (see Chapter 28).
Different groups of 1s and 0s then represent different letters in the secret
message. A Tudor master of spycraft, Sir Francis Bacon, came up with
the original version of this kind of steganography. His version used capital
and lower case letters for the two binary symbols in writing that was
apparently about something completely different.
 Sniff Out that Card: Steganography 301

In the trick, one bit of binary information (a single 1 or 0) was stored

in the image on the back of each card, based on which way up it was.
The right way up (0) meant “I am not the chosen card”. The wrong way
up (1) meant “I am the chosen card”.
You could store even more information in a pack of cards in this way.
For example, take a pack of cards and turn each card the right way up
(meaning 0) or the wrong way up (meaning 1) to spell out a message
in binary through the whole pack. Every group of five cards would then
represent a letter of the alphabet running from 00001 meaning A to
11010 meaning Z. Send the pack of cards to someone supposedly as
a present but where the real reason is the message you are sending
them. Of course, one pack of cards could only hold a 10-letter message
this way.
In the computer age, there are all sorts of ways to hide messages.
One way is in a photo. Digital images are ultimately stored as binary, so
binary can be hidden there. An image is split into a grid of tiny squares,
called pixels (see Chapter 29), and the colour of each is stored as a
binary code. The more binary bits used for each pixel, the more colours
can be used.
Now imagine an image of a beach with a blue sky with wispy clouds.
Perhaps the sky uses two shades of blue, a deeper blue and a lighter
blue. Suppose the difference between the binary version of the two
colours is one binary bit. If that bit is a 0, then the lighter blue is
used there. If it is a 1, then the darker blue is used. A message could be
hidden by setting the blue of specific pixels (perhaps every 10th pixel)
based on the binary of the message to be stored. If the image is of
high resolution and so has thousands of pixels in each row, this will
be hard to spot unless you know it is there, and a fairly long message
might be hidden. Of course, binary can represent anything so what is
hidden could even be a different smaller image spelled out in the changed
1s and 0s.
302 Conjuring with Computation

PAUSE: Sir Francis Bacon

Tudor philosopher, Sir Francis Bacon (1561–1626) is best known
for pioneering the scientific method. He was also a Tudor master of
spycraft. He invented a form of steganography that was the first use
of a binary system for encoding letters. Each letter in a message was
represented by a different sequence of five As or Bs. These were then
hidden in writing that was apparently about something completely
different, using other things in place of A and B. There are various
ways this can be done. He used one font for A and a different font
for B letters. Another simple version he used was capital letters for
A and lower case letters for B. This also demonstrates how any
symbols can be used for 0 and 1 in binary.
 Chapter 43

Classic Mentalism
Codes and Ciphers

Conjuring
One of two magicians leaves the room while a card is chosen. When
they return, they instantly announce the card that was chosen.

Computation
Cryptography, the use of codes and ciphers, to make messages
unreadable, is a core part of how computers are made secure.
Used by rulers, the military and spies for thousands of years to
keep their secrets safe, it is now the way banks can send digital
money around the world, how we keep documents, chat and video
meetings secure as well as the basis for preventing anyone logging
into our computers.

The Trick
In this trick, there are two magicians. One stays with the audience, while
the other, “the receiving magician”, leaves the room. A brand new pack
of cards is opened. A member of the audience chooses a card from the
pack. The chosen card is placed back in the pack. The pack is put back

303
304 Conjuring with Computation

in the box and placed back on the table. The receiver is called back.
When they return, they immediately tell the audience the chosen card.

How It Works
The secret to this trick is in the table. Once the card has been selected,
pack and box are placed on the table in a position that depends on the
card selected. The two magicians agree on a secret code of where the
box will be placed in advance such as (see Figure 43.1):
If the box is face up and its flap is closed: Clubs.
If the box is face up and its flap is open: Spades.
If the box is face down and its flap is closed: Diamonds.
If the box is face down and its flap is open: Hearts.
That’s the suits taken care of. Now for the value. The performers again
agree in advance, this time on how to chop up the card table into mental
zones. Horizontally, you have the top row, centre row and bottom row
of the table and far right, mid right, mid left and far left. That’s 3 × 4
unique locations. You have 12 places for 12 of the 13 card values. Ace
is top-far-left. Two is top-middle-left, and so on (see Figure 43.2). The
magician places the pack of cards in the correct pre-agreed position, box
face up or not, flap open or closed as needed (see Figure 43.3).

Figure 43.1. The suit is determined by whether the box is left face up or face down
and with the flap open or closed.
 Classic Mentalism: Codes and Ciphers 305

Figure 43.2. Divide the table into mental areas like this to indicate what the card
value is. If the card is a King, put the box in the volunteer’s hand.

Figure 43.3. The cards are in the top row, far right (4), and the box is face up, with
flap open (SPADES), so card indicated is the 4 of spades.
306 Conjuring with Computation

What about the 13th possibility? That is when the card is the King.
If the card isn’t the King (most often), you leave the pack of cards on
the table. If it is the King, have your volunteer place their hand out flat
and leave the box of cards on it. Ask them to concentrate on the card
they chose while staring at the pack.

The Magical Algorithm

TO DO Classic Mentalism:
1. The receiver leaves the room.
2. A new pack of cards is opened.
3. A volunteer chooses a card and shows it to you and the audience.
4. They place the card back in the pack.
5. You put the pack back in the box.
6. Orient the box using the Suit Code to encode the suit as an
orientation.
7. Place the box using the Face Value Code to encode the face
value as a position.
8. Call the receiver back in.
9. The receiver works out the suit using the Suit Code Table to
decode.
10. The receiver works out the face value using the Face Value Code
Table to decode.
11. The receiver names that face value and that suit.
To encode, read the code tables left to right. To decode, read the code
tables right to left.

Suit Code Table:

CLUBS: face-up-flap-closed.
SPADES: face-up-flap-open.
DIAMONDS: face-down-flap-closed.
HEARTS: face-down-flap-open.
 Classic Mentalism: Codes and Ciphers 307

Face Value Code Table:

ACE: top-far-left.
TWO: top-middle-left.
THREE: top-middle-right.
FOUR: top-far-right.
FIVE: middle-far-left.
SIX: middle-middle-left.
SEVEN: middle-middle-right.
EIGHT: middle-far-right.
NINE: bottom-far-left.
TEN: bottom-middle-left.
JACK: bottom-middle-right.
QUEEN: bottom-far-right.
KING: hand.

Thinking Computationally
This trick is using steganography again. However, it is also using a secret
code. In a simple version of steganography, a message is just hidden (like
growing hair over it or hiding a letter in a waterproof package inside a
barrel of ale). In more complex versions as in both the last trick and this
one, information is transformed into a different form that also hides it
using a secret code.
A code is just an algorithm that transforms information from one
form into another. Encoding transforms the information and decoding
transforms it back to the original form (see Chapter 25). The information
encoded can be in any form, whether spoken, written, gestures or implicit
in the form of objects. The decoded version of the information can also
have any form. For example, a semaphore encodes written or spoken
letters into the positions of a person’s arms holding flags. A secret code
is one where the code is being used in a secret way. Only some people
know the secret, know the code.
In the last trick, we had a very simple encoding of one piece of
information “It is this card” into the orientation of the card. In this trick,
308 Conjuring with Computation

we use a code again, but a more complicated one that encodes more
information: that of the details of one out of 52 cards. Separate codes
are used for the suit and for the value of the card. Codes like this are
the basis of a wide variety of mentalism acts. They only appear magical
because they are both secret and hidden. For example, in an act where
the magician takes an object from an audience member and their partner
blindfolded can say what it is, different words the magician says in asking
them what they are holding might mean, “It’s a watch”, “It’s a mirror”,
and so on . . .
Computer scientists also use a wide variety of codes to make
information easier to store and transmit. Most codes are not secret so
are not a secure way to keep information secret. However, secret codes
have also been used throughout history to keep information secure. This
is known as cryptography. Even if the message is intercepted, if the
enemy can not decode it, they cannot read it. They may though be
able to infer something useful just from the fact that it is written in
code, who it is from and who it is to. Magicians use codes in this way,
usually combined with steganography, explicitly to also hide the fact
that information is being passed. Many of the tricks in this book have
used information secretly both hidden and encoded one way or another.
The word code, in general use, means any kind of secret message.
In cryptography, it means a specific kind of encryption algorithm where
a word or phrase in the message is allocated a code word, symbol or
phrase. Mentioning, as part of a conversation, the code word “Phoenix”
might mean “Attack at dawn”, whereas “Vulture” means “The attack is
aborted”. That is what our trick did. The word “Clubs” is given a code
word “face-up-flap-open” which is then translated into the corresponding
physical positioning: the box is placed face up with its flap open. What
we do not do is give the separate letters C-L-U-B-S their own code word
and try and spell them out in code. Native American languages were
used as codes in World War II as very few people spoke them. Navajo
words were used as code words for their English equivalent.
A cipher, by contrast, works on individual letters (or syllables).
A simple cipher encrypts each letter to a new symbol or symbols to
encrypt a message. Modern encryption generally uses ciphers rather than
codes as it is more flexible. The Caesar cipher is a simple example
 Classic Mentalism: Codes and Ciphers 309

used since Roman times. Each letter of the alphabet is replaced by

the letter three places earlier in the alphabet. Today, complex ciphers
are used in all sorts of ways, from password-protecting your documents
to securing instant messaging conversations and transferring money
securely between banks and businesses across the Internet.

Interlude: Reginald Scot

The first English book giving details of conjuring tricks was The
Discoverie of Witchcraft by a member of parliament, Reginald Scot
(c.1538–1599), published in 1584. His aim in the book as a whole
was to prove by reason that witches did not exist. This was in the
midst of witch hunts that were persecuting thousands of innocent
women across the country. His section on conjuring aimed to debunk
charlatans. It included a way to communicate secretly by code with
a confederate from behind a door. The confederate would indicate
whether coins have been arranged in a pile or as a cross, using a
simple code: the difference between him saying “What is it?” versus
“What ist?” The book also included a trick of how to appear to chop
off your head. Shakespeare supposedly used the book for inspiration
about the witches in Macbeth.
This page intentionally left blank
 Chapter 44

Call the Clairvoyant

Legitimate Channels

Conjuring
A friend picks a card from a normal pack of cards. You phone the
mysterious “Clairvoyant”. The “Clairvoyant” is able to name the
chosen card.

Computation
Computer systems are kept secure by having separate areas with
different security levels. Nothing from a high-security area should
be able to leak to a lower-security area. Hiding information
(steganography) in a legitimate channel (an allowed way to com-
municate) is one-way information that might be leaked. Hackers
often exploit legitimate channels in computer systems. Magicians
build tricks around them too.

The Trick
This is a trick that could be done with an audience or just with a friend.
You announce that you know an amazing person who goes by the name

311
312 Conjuring with Computation

“The Clairvoyant” who can do amazing things and if you like you will
ask them to do a demonstration of their powers. Close any doors and
curtains of the room you are in so no one outside can see in. Next,
ask the friend (or a volunteer) to choose any card from a face-up pack,
allowing everyone, you included, to see the card they choose. Next phone
the Clairvoyant. When they answer you simply say, “I have a call for the
Clairvoyant”. After listening for a moment, you say “Thanks”. After a
further pause during which you say nothing, you finally say, “Let me pass
you over.” You hand the phone over to the friend/volunteer (and/or put
the phone on speaker phone so the audience can hear) telling them
to just say “Hello” to the Clairvoyant and then listen. They speak to
the Clairvoyant. Despite the Clairvoyant not being in the room, having
apparently no way to know what happened, and you saying nothing to
indicate what the card was, they dramatically tell your friend the card
they chose . . . and they are right.

How It Works
When you first phone, your accomplice, the Clairvoyant, just names the
card suits, saying: “Clubs . . . Spades . . . Diamonds . . . Hearts”. You say
“Thanks” when they reach the suit of the chosen card (which you, but
not they, have seen). Your phoney Clairvoyant accomplice now knows
the suit and starts counting out the values, Ace to King. Again you stay
silent until they get to the chosen card, at which point you this time
say: “Let me pass you over.” Figure 44.1 shows an example of how the
5 of spades is communicated. Your accomplice now knows both suit
and value. Once they are passed to the person who chose the card, they

Figure 44.1. Communicating the 5 of spades to the Clairvoyant. The Clairvoyant

lists the possibilities until interrupted by the magician speaking.
 Call the Clairvoyant: Legitimate Channels 313

know the suit and value and so can surprise them by immediately telling
them what it was.

The Magical Algorithm

This trick involves two algorithms that work together to allow informa-
tion to be communicated.
The first, followed by the Clairvoyant, the receiver of the information,
is:
TO DO Call the Clairvoyant, the Clairvoyant must:
1. Answer the phone when it rings.
2. IF the person phoning says “I have a call for the Clairvoyant,”
THEN
a. Name the suits slowly “Clubs, Spades, Diamonds, Hearts”,
stopping when the other person says “Thanks”, writing the
last suit spoken down.
b. Name the face value of cards slowly “Ace, Two . . . Queen,
King”, stopping when the other person says “Let me pass you
over,” writing the last value spoken down.
c. When they hear the second person, say “Hello”, they say “Your
card was . . .”, and name the value and suit they wrote down.
The second algorithm followed by the magician is:
TO DO Call the Clairvoyant, the magician must:
1. Tell the friend about the Clairvoyant.
2. Close doors and shut curtains so the room is secure.
3. Ask the friend to pick a card, showing you what it was.
4. Phone the Clairvoyant (the person who has agreed to follow the
first algorithm).
5. Say “I have a call for the Clairvoyant.”
6. Listen to them saying the suits and as soon as the chosen suit is
spoken, say “Thanks”.
7. Listen to them saying the values and as soon as the chosen value
is spoken, say “Let me pass you over.”
8. Hand over the phone to the friend, telling them to say “Hello”
and then listen to what the Clairvoyant has to say.
314 Conjuring with Computation

9. They say “Hello” and listen to the Clairvoyant naming their

card.

Thinking Computationally
This trick requires a shared understanding that the two algorithms will
be followed. Such a pair of algorithms used to pass information is called
a communication protocol. Even though you apparently say nothing to
give away what the card was (and in fact say the same thing every time
you do the trick), you do pass the information on to your accomplice.
You do it not by what you say but in the timing of when you say it.
Your silence passes information!
Computers that are intended to be secure use secure operating
systems: software that controls the computer but does so in a way
that is supposed to not allow information to be passed around. For
example, it may be divided into high-security “Top Secret” and low-
security “Unclassified” areas. Top-secret information is stored in the
high-security areas which only certain people can access. Meanwhile,
everyone else has access to the “Unclassified” areas where only non-
confidential material resides.
In our trick, think of the room where you, the magician and your
friend are, as being the top-secret area. That is where the chosen card
resides. You are both there so know the secret information, that is the
chosen card. The Clairvoyant, on the other hand, is elsewhere. They are
in an unclassified area.
An important part of the setup for the trick is that they should
clearly not be able to just see the information (the card) that resides in
the top-secret area. This is clear because they are not in the room, just
on the end of the phone, and you have closed all doors and curtains.
Your accomplice cannot see in! Let us call that BELIEF 1.

Legitimate channels
The only way information can pass out of the top-secret area to the
unclassified area is therefore via the phone. This is what a computer
 Call the Clairvoyant: Legitimate Channels 315

scientist would call a Legitimate Communication Channel: one built

into the system. However, it is monitored; the volunteer and audience
hear everything you say, so they apparently hear all information being
passed out of the area. Think of this as BELIEF 2. You apparently cannot
pass information out. (Except you can!).
For the trick to seem magical, the audience have to hold both beliefs.
It works because you do have a secret, undetectable way of passing the
card details out undetected: the secret pair of algorithms allows the top-
secret information to pass out of the top-secret area and be written down
in the unclassified area! It is using the legitimate channel of the phone
call, but in a hidden way (so using steganography, see Chapter 42).
Secure computer systems are split into areas with different security
levels like our top-secret and unclassified areas, and there are two
important rules built into secure systems that match the two beliefs
behind our trick:

1. Those in lower-security areas (like the unclassified area) should

not be able to read any information in higher-security areas (the
top-secret area).
2. Someone with access to the top-secret area should not be able
to pass information out (accidentally or maliciously) into the
unclassified area. The security system should not allow it. Some
people, however, may be given the status of being trusted: they
are able to pass information out but are trusted not to do so with
information they should not.

The magician is acting as a Trusted Person here. They could actually

communicate information out, but everything they say is monitored, so
you are led to believe that you would know if they did. The flaw in the
security here is that you are not monitoring what the other person is
saying.
The typical spy and whistleblower scenarios, where a trusted person
inside an organisation passes information out, are working in this way.
For example, Edward Snowden, as part of his job, was given, copied and
leaked hundreds of thousands of classified US documents to journalists
316 Conjuring with Computation

in 2013. These revealed to the public the massive extent to which the
US was spying on them, in breach of the US constitution.
Snowden took the information with him which is a bit like our
magician leaving the room having seen the card (not very magical).
In the trick, we are using steganography to hide the information being
leaked in plain sight through the legitimate channel. The message
about which card is chosen is passed in an apparently run-of-the-mill
conversation. It is not what is said that communicates the information
but the timing of when it is said that matters.
In fact, in a more sophisticated variation of the trick, the magician
and the Clairvoyant could pass the information without the Clairvoyant
saying anything or saying something innocuous, and instead just timing
the gap between the magician speaking. This would work as long as the
pair had agreed in advance the timings of each suit and face values. So
a pause of 5 seconds from saying “I have a call for the Clairvoyant” to
saying “Thanks” means Clubs. A pause of 10 seconds means Spades,
and so on. Then the whole conversation could be on speakerphone the
whole time, and the information still passed.
Timing can similarly be used by a program to communicate covertly
out of a secure network. Information might be communicated by the time
a message is sent rather than its contents, for example. Alternatively,
information might be passed based on whether a particular encrypted file
exists or not. They cannot be read, but they can still leak information as
a program on the inside could create and delete files to communicate.
Making a system really secure is actually quite tricky!
 Call the Clairvoyant: Legitimate Channels 317

Interlude: Sydney and

Lesley Piddington
In the 1950s, Australian couple Sydney and Lesley Piddington took
the entertainment world by storm. They had the nation perplexed,
puzzled and entertained. They were seemingly able to communicate
telepathically over great distances. It all started in World War II when
Sydney (1918–1991) was a prisoner of war. To keep up morale, he
devised a mentalism act where he “read the minds” of other soldiers.
When he later married Lesley (1925–2016), they perfected the act
and became an overnight sensation, attracting BBC radio audiences
of 20 million listeners.
They communicated random words and objects selected by the
audience, even when Lesley was in a circling aeroplane or Sydney was
in a diving bell in a swimming pool. To this day, their secret remains
unknown, though many have tried to work it out. Perhaps they used
a hidden transmitter. After all, that was fairly new technology then.
Or perhaps they were using their own version of the old mentalism
trick: a code to transmit information hidden in plain sight.
Sydney had a severe stutter, and some suggested it was
the pauses he made in words rather than the words themselves
that conveyed the information. Using timing and silence to code
information may seem rather odd, but it can be used to great effect
both in tricks and espionage and has been suggested as a way to
hide messages in video calls, for example.
This page intentionally left blank
 Chapter 45

Call the Clairvoyant a Second Time

Covert Channels
and the Surveillance Society

Conjuring
The Clairvoyant is called once more, this time able to tell exactly
which object from a table full of objects a member of the audience
chose even though called on a speakerphone, so all could hear
everything that was said.

Computation
Another way hackers do damage is to set up covert channels: ways
of communicating that are not supposed to be there. Magicians
build tricks around covert channels too.

How to Do the Trick

Explain (as in the previous trick) that you know someone with amazing
powers who goes by the name “The Clairvoyant” and you will ask them

319
320 Conjuring with Computation

to do a demonstration. Arrange on a table a range of different everyday

objects, perhaps borrowed from different members of the audience,
together with some magical objects you provide: a pen, a watch, a coin,
a silk handkerchief, a wand, a top hat, a pack of cards, and so on. Shut
curtains and doors to ensure no one can see in.
Have a member of the audience study the objects and choose one,
pick it up, and show it to the audience and then ask the whole audience
to think hard about that object only.
You then phone the Clairvoyant using a speakerphone. When they
answer, you simply say “I have a call for the Clairvoyant.” They say in
a mysterious voice (perhaps covering the phone with a handkerchief)
“I am she.” You then ask them “I have a group of people with me
thinking of a single object. Can you say what that object is?” After a
silence of several seconds, they correctly announce what the object was.

How It Works
You simply set up a small but hidden webcam somewhere in the room
where it has a clear view of the objects on the table, or of where the
volunteer will stand holding the chosen object (perhaps poking through
curtains at the side of the stage). The webcam livestreams the images
so that the person at the other end can see everything. This could just
be done by setting up a video call to the computer the webcam is linked
to. The Clairvoyant watches the show via the video call (with the sound
off!) and so sees the object taken.
When internet-based video calls were new, famous magicians used
such technological variations of this. Given such video surveillance
is now totally pervasive, making this magical today will depend on
your presentation skills distracting the audience from thinking it is a
possibility!

The Magical Algorithm

This trick again involves two algorithms that work together to allow
information to be communicated.
 Call the Clairvoyant a Second Time 321

The first, followed by the Clairvoyant, the receiver of the infor-

mation, is:
TO DO Call the Clairvoyant a Second Time, the Clairvoyant must:
1. Watch the video stream until the volunteer picks an object.
2. Answer the phone when it rings.
3. Say in a mysterious voice: “I am she” (or he as appropriate).
4. Pause.
5. Name the object that was chosen.
The second algorithm, followed by the magician, is:
TO DO Call the Clairvoyant a Second Time, the magician must:
1. Tell the audience about the Clairvoyant.
2. Close doors and shut curtains so the room is secure.
3. Lay out a mixture of normal and magical objects on the table.
4. Ask a volunteer to pick an object, showing it to the audience.
5. Have the whole audience concentrate on the object, thinking only
about it.
6. Phone the Clairvoyant (the person who has agreed to follow the
first algorithm) putting the phone on speakerphone.
7. Say “I have a call for the Clairvoyant.”
8. Say “I have a group of people with me thinking of a single object.
Can you say what that object is?”
9. Listen to their answer and thank them, announcing that they were
correct.

Thinking Computationally
Covert channels
The trick is similar to the last though done in a different hi-tech way,
using the hidden webcam in the room that was positioned to be able to
see the object so the Clairvoyant can watch proceedings via the camera
streaming images from the room. For the trick to work, the presence
of the hidden camera needs to be unknown to the audience (unlike the
phone connection which everyone knows exists). The phone connection
to the Clairvoyant is a legitimate channel. In this trick, by contrast, we
322 Conjuring with Computation

are using a covert channel: a way of communicating that should not

exist and is not known to exist by anyone but those using it.
Hackers create covert channels as a way of gaining information too.
For example, if you visit a dodgy website, and allow it to execute code
on your phone or computer, then the website might insert a program
(a trojan horse program) that is unknown to you. It records and sends
to the hacker details of all your keypresses, say, or allows them to switch
on your phone’s mike or camera so they can see or hear what you do.
This would be a covert channel leaking information from your computer
or phone to the hacker’s computer, just as the webcam in this alternative
version of the trick does.

Surveillance society
The trick depends on camera surveillance. This has now become
ubiquitous in many western countries. As you walk through the centre
of a city like London, cameras are everywhere mounted on buildings.
You are potentially constantly being watched. As you drive around the
motorway system, your number plate is constantly being logged. As
you move anywhere, your phone logs your position as you connect to
different phone masts. This is an example of technology being used for
good purposes that could easily be subverted, whether by individuals
such as stalkers, companies or states. More recently, companies selling
hacking tools to countries have provided them with ways of taking over
a person’s phone including the microphone and camera, so they can
see and hear everything you can. Supposedly only used for legitimate
policing, it has been used repeatedly to monitor political opponents
of oppressive states, legitimate journalists, lawyers and more, in clearly
unethical ways.
Privacy can no longer be taken for granted.
 Call the Clairvoyant a Second Time 323

PAUSE: Hedy Lamarr

Hollywood actress, Hedy Lamarr (1914–2000), was a film superstar
playing parts such as Helen of Troy, Joan of Arc and Delilah. Most
definitely not just a pretty face, she was also an inventor. Working
with film music composer, George Antheil at the start of World
War II, she came up with a way to guide radio-controlled torpedoes
to their targets. The idea was ignored at the time but is now an
absolutely central part of wireless communications. We now call it
frequency hopping: the idea is that rather than sending out control
signals on one frequency, the frequency constantly changes. Only
the transmitter and receiver know the frequency it will jump to next
(stored on a piano roll in their design). This made it hard both to
intercept the radio signals and to jam them: exactly what was needed
for torpedoes had it been taken up.
This page intentionally left blank
 Chapter 46

Reading a Personality with

Hot Chocolate
Big Data and Privacy

Conjuring
You make a mug of hot chocolate. The person whose personality
you will read stirs it five times and then drinks. You stare into the
sludge of chocolate left behind and from it read the personality of
the volunteer.

Computation
Tech companies collect data about everyone who uses their
services so they can use the information to sell advertising and
target services at you. They build up a very detailed picture of
each person, their personality and their lives. Their aim is to know
you better than you know yourself.

The Trick
You explain that some fortune tellers can read people by looking in a
crystal ball, others in tea leaves. You, however, having spent so much

325
326 Conjuring with Computation

time studying chocolate, can tell a person’s personality by looking at the

dregs left behind when they drink hot chocolate. You make a small mug
of hot chocolate. The person stirs it five times and then drinks it. Why
five times? Five is a special magical number, the number of pentagons
and pentagrams. Once they have finished, you stare into the empty mug
at the undrunk chocolate sludge stuck to the bottom.
After a few moments, you tell them something about their person-
ality,

At times, you are overly critical of yourself, but you also have a
deep desire for others to appreciate you. Sometimes, you are very
confident but there are times when you are less sure of yourself.
You often worry about whether you have made the right decision
and some of the things you try to do do not work out. You like
variety and dislike it when prevented from doing things because
of rules imposed by others. Sometimes you are sociable, but you
can also be much more reserved. You have occasionally found
it to be a bad idea to reveal too much of yourself to others too
freely. In general, you are an independent thinker who doesn’t
just believe things you are told without some kind of evidence.
Most of the time, you can overcome the flaws in your personality.

Ask them to rate how accurate your description is, and whether they
recognise themselves.

How It Works
You just say the same thing whoever the person and whatever the state
of the chocolate mug, occasionally poking at the chocolate as though it
is revealing new facts.
The effect used in the trick is called the Barnum–Forer effect after
showman Phineas T. Barnum and psychologist Bertram R. Forer. It is
relied on by the likes of astrologers and fortune tellers.
The personality prediction is full of the same kind of statements
you find in horoscopes, called Barnum statements. Those above are
 Reading a Personality with Hot Chocolate: Big Data and Privacy 327

variations on ones used in a classic experiment by psychologist Bertram

R. Forer which demonstrated the effect. In 1948, he gave a “psychology
test” to his students and then gave each student what he claimed was
a personalised summary of their personality based on the test. He asked
them to rate the summary, and on average, they rated them at over
4 on a scale of 0–5 suggesting they thought they were very accurate
summaries of their personality. In fact, he gave them all an identical
summary based on the same statements that he had pulled from an
astrology book.
Barnum statements aim to be very general statements that sound
specific but that lots of people would agree with. People can apply their
own meaning to them, matching them to things they believe about
themselves. Some of the statements include phrases like “At times”
or “Sometimes” that mean they are likely to apply to anyone. Whilst
overall they should be mainly positive traits (as people are more likely to
accept positive statements about themselves), they also mix positive and
negative statements about the same thing: saying things like “Sometimes
you are confident but sometimes you aren’t”, that are just bound to be
true, though saying it in a more long-winded way. It is also important
that the person believes the summary is personal to them and that they
trust in the authority of the person doing the test.

The Magical Algorithm

TO PREPARE Reading a Personality:
1. Memorise a series of Barnum statements such as those in the given
paragraph.

TO DO Reading a Personality:
1. Tell the person you can read their personality in the chocolate
dregs of a mug of hot chocolate.
2. Make them a mug of hot chocolate.
3. They stir the drink five times.
4. They drink the hot chocolate.
5. You stare into the chocolate sludge left behind.
328 Conjuring with Computation

6. You “describe their personality” while staring into the mug, poking
at the sludge with a teaspoon occasionally and saying a series of
Barnum statements.
7. Ask them to rate your accuracy on a scale of 1–5 with 5 meaning
fully accurate.

Thinking Computationally
Fortune tellers have long used this kind of fraudulent reading to take
people’s money. Oppressive regimes have also long collected information
dossiers on their citizens so they know who is likely to oppose them.
Hackers and fraudsters also use social engineering to get people
unwittingly to give up personal information, pretending to be people
they aren’t for example and asking for passwords and credit card details.
Magicians have also used a version where they just ask for information
for mentalism and spiritualism acts. One version involves collecting
information about their audience members before the show to use to
make predictions seem more realistic.
Now technology allows it to be done for real on a mass scale. We
blithely give away lots of information about ourselves to technology
companies in exchange for their services in a way that allows them to
both know and predict a lot about us. The way we follow other people on
social media combined with our “likes” can be used to fairly accurately
tell things like our ethnicity and gender and who we are likely to vote for,
for example. Companies also collect your browsing history, what kind of
music you listen to and films you watch as well as what you buy or think
of buying. Technology companies sell the data to marketing companies
who then compile very detailed sets of statements about every one of
us based on the mass of information collected and combined. These are
no longer Barnum statements but accurate guesses about very intimate
aspects of your personality: interests, habits and life.
All this, once private, information can be used for good or bad.
It can be used to make life more convenient but also to manipulate
us. Advertisers use the information to sell you things you probably don’t
need or really want, political parties use it to target messages suggesting
they support the things that matter to you to change the way you vote,
 Reading a Personality with Hot Chocolate: Big Data and Privacy 329

and other companies use it to run promotional campaigns to change

your opinion so you think they are good companies, or to change your
mind about issues that further their aims. Governments use it to promote
their messages and sway opinion of those in other countries in ways that
favour them and undermine democracy. Some governments use it to
identify people likely to oppose them.
It is done using big data processing. Similar data is collected about
large numbers of people, and general patterns are gleaned. The more
data collected, the more detailed the patterns that can be determined.
Machine learning algorithms are used to spot the patterns and then
apply them. They are artificial intelligence algorithms, partly modelled
on the way our brains work, but all they really do is find patterns in their
vast store of data. They then look for matches to those patterns in new
data, about a new person, say, using the links found to form conclusions
about that person.
People who watch particular news stations and follow particular
people on social media have been found to be grouped with people who
believe abortion is bad, and also those who vote a particular way. If a new
person is found to watch that station and follow the same people, then
it can be predicted they believe abortion is bad and vote the same way
too. If messages are sent to those people claiming an election candidate
is pro-abortion (whether they are or not), then that may help swing
the vote. Meanwhile, a different group of people identified as being
interested in the environment because of programs they have watched
are sent messages suggesting that the politician being promoted cares
deeply about green issues (whether they do or not).
The predictions are far more powerful when combined with actual
data about whatever an organisation wants to predict. Many people
make such data freely available: perhaps they both give facts about
themselves and state their beliefs about the environment in their social
network profile, for example. That hard data can be used as the basis
for predicting private information of others. One ruse is to use bogus
“surveys” to ask people for specific information. This known information
then ties patterns to specific beliefs and personality traits. It can then be
used to predict that others that have been grouped with those people
will be similar in other aspects of the grouping detected. It therefore
330 Conjuring with Computation

gives away similar information about lots of people who refuse to fill in
the surveys because they prefer to be private. That freedom of privacy
is effectively removed from them by the choices of others.
This kind of manipulation of messaging can be helpful, for example,
if it helps people understand the true beliefs of the politicians being
promoted. On the other hand, if it is used to propagate lies and half-
truths, it undermines democracy.
Computer science is not just about technology but also about its
place in society. New technology changes the way we do things, but
also can completely change the way the world works, for good or
bad. Understanding the effects of changes on society is important, and
underlying this is also an understanding of, and commitment to, using
technology for good, not just personal gain.

Interlude: Phineas T. Barnum

The kind of statements used in the “Reading a personality with
hot chocolate” trick is called Barnum statements because showman
Phineas T. Barnum (1810–1891) is credited as having used them.
He was happy to use effects and tricks to entertain. However, he
went out of his way to debunk fraudsters who used the same tricks
to actively con people. This included those who tried to convince
others of their “psychic” abilities or that they could “speak to the
dead”, using the tricks as a way of extracting money from the
newly bereaved. He testified at the trial of one fraudster who used
photos that had ghostly images of dead relatives in the background.
Barnum paid a photographer to create a picture of himself with a
ghostly Abraham Lincoln in the background to show how easily spirit
photos could be faked. This was long before computer software like
Photoshop made it easy for anyone to do this kind of fakery with
images.
 Chapter 47

The Chevreul Pendulum

Professional Ethics

Conjuring
A volunteer holds a pendulum over a circle with YES and NO
written on it. They use it to “talk to the spirit world” and ask
questions about their future.

Computation
Professional ethics is very important for all computer scientists.
A person with computational thinking skills and computing knowl-
edge can use it for good or bad: as a black hat hacker breaking
into systems stealing data or as a white hat hacker helping protect
people’s privacy, for example. Big data can be used to spy on
people or to improve their lives. Behaving ethically matters.

The Trick
As part of a magic show, ask for a volunteer who is open to the idea that
the spirit world might exist. Tell them they will take part in a seance
where they will talk to a supernatural Victorian spirit you have made

331
332 Conjuring with Computation

contact with, called Alice. You know Alice likes to communicate in this
way and likes to talk about relationships whether past, present or future.
Give the volunteer a simple pendulum made from a large, heavy, old-
style key hanging on the end of a string 40 centimetres long. Explain that
the key is so old and has been around humans and their lives so long,
and it has become a link to the spirit world. Have the volunteer hold it
over a large circle split into quarters with YES written in two opposite
quarters and NO written in the other two quarters (see Figure 47.1).
You may wish also to decorate it with runes or other ghostly images.
The volunteer should be standing and keep their arm held out straight
and unsupported. They should hold it so it hangs over the middle of the
circle. Tell them they must concentrate on the pendulum at all times,
but not try to consciously move it themselves. They should allow Alice
to work through them.
You then ask a series of questions, starting with:

Alice, are you there? . . . we wish to communicate, are you willing

to answer our question?

Follow this by the first question. If after a short wait the pendulum starts
to swing to yes or no, then announce the answer given. Then ask the

Figure 47.1. The target for the Chevreul Pendulum.

 The Chevreul Pendulum: Professional Ethics 333

person holding the pendulum (or their friends) to come up with further
questions for Alice to answer that they would like to know the answer
to, whether about past, present or future relationships. If they are open
to the influence of spirits, you say, then the pendulum will swing to give
answers to the questions.

The Magical Algorithm

TO PREPARE the Chevreul Pendulum:
1. Make a pendulum by tying an old key on the end of a
40-centimetre long string.
2. Draw a target: a circle split into quarters with YES written on one
pair of opposite quarters and NO on the other pair.

TO DO the Chevreul Pendulum:

1. A volunteer holds the pendulum with their arm held out so it is
above the centre of the target.
2. Tell them to concentrate on the pendulum but not move it
consciously themselves.
3. Ask the volunteer for a YES/NO question they would like Alice
to answer about the past, present or future.
4. DO THE FOLLOWING UNTIL the audience do not have any more
questions:
a. Say “Alice, are you there? . . . we wish to communicate, are you
willing to answer our question?”
b. Repeat the question.
c. Watch to see if the pendulum starts to swing showing Alice is
communicating through the volunteer.
d. IF the pendulum swings,
THEN announce the answer indicated,
OTHERWISE announce that you will have to take it as an
“Unknown . . . even the spirits cannot tell.”
e. Ask the audience if they would like to know more and if so to
give you another YES/NO question they would like Alice to
answer about the past, present or future.
334 Conjuring with Computation

How It Works
This illusion is based on an effect called an ideomotor response
and named after Michel Eugene Chevreul, a French scientist who did
experiments about the effect in the 1830s. The same effect leads to
the movement of tables and Ouija boards in seances. The movements
happen because ideas can lead to subconscious physical changes. An
example is the way your mouth waters when you think of something
you like such as chocolate (for me at least). In the trick, thinking
about questions leads to very small movements of the hand of the
person holding the pendulum. The pendulum magnifies the movements
particularly if the arm is moving backwards and forwards in a way that
matches the natural frequency of the pendulum swinging (which depends
on its length). Staring at the pendulum increases the effect as the person
thinking strongly about the pendulum swinging one way or the other can
trigger the subconscious movements that cause it to happen.

Thinking Computationally
Whilst it is fine to do a trick like this as part of a magic show where
it is clearly intended to be just for fun, it would be unethical to use it
to really claim to talk to the dead relatives of people as you are not. It
would be even more unethical to charge people to do so.
Spiritualism became really popular in Victorian times where seances
were held to allow people to supposedly speak to the dead. Whilst a
few may have truly believed they were helping people talk to their dead
relatives, most were just frauds, unethically preying on people who had
lost loved ones. Their aim was to swindle money from them using their
false claims. Other people use the same sort of tricks as magicians for
a variety of unethical reasons: to cheat when gambling, to claim to do
faith healing and to con people out of money.
Behaving ethically is very important to most magicians. They make
a clear distinction that what they are doing is a performance for
entertainment. Some go out of their way to expose frauds who use the
kind of illusions and conjuring tricks we’ve seen in this book to fleece
people by claiming them to be something other than mere tricks that
are part of a conjuring show.
 The Chevreul Pendulum: Professional Ethics 335

Professional ethics is also very important for computer scientists.

Those with computing skills can use them for good or bad, and it is
important that people do behave ethically. Just as magicians can use
their skills to devise tricks to present as other things to fool people, so
programmers can devise programs that pretend to be one thing while
being something else. Such malware is developed and used by hackers
to, for example, steal data for gain, steal someone’s identity to defraud
them, or lock people out of their computers until they pay a ransom.
The computing skills being used are not in themselves bad, it is how
they are used that is bad. A magician charging an audience member to
see them do tricks that is clearly presented as a night of entertainment
is not the same as someone using tricks to convince people they are in
contact with gods or ghosts and so take money as a collection “for the
work of God” or so they can “speak to dead loved ones”. Similarly, those
with hacking skills can use them for good or bad. For example, some
companies employ white hat hackers to search for problems with their
security systems so they can make them more secure. They use the same
skills as the bad people, also earning money, but in an ethical way.

Interlude: James Randi

James Randi (1928–2020) was a magician famous for investigating
fraudulent claims, having shown how a whole series of people
claiming to have supernatural powers were actually just using magic
tricks. He also demonstrated flaws in research projects supposedly
investigating the paranormal. He has shown how claim after claim
of paranormal ability was easily done with magic tricks.
This page intentionally left blank
 Part 12

Advanced Technology

We have mainly focussed on the links between magic and computer

science ideas rather than gadgets, but there is a long history of magic
tricks being based around novel gadgets based on new science or
engineering developments.
This page intentionally left blank
 Chapter 48

The Faerie Cage

Augmented Reality

Conjuring
You show the audience a faerie you have trapped in a magical
cage.

Computation
Virtual reality involves creating online worlds that people can
move digital versions of themselves around in. Augmented reality
involves superimposing virtual elements on the real world so that
you can move around both real and virtual worlds together and
so that nonexistent things appear to be there in the real world.
Applications range from games to the head-up displays in fighter
aircraft.

The Trick
Explain that faeries do exist and you have trapped one inside a special
box lined with rowan wood, which acts as a barrier to a faerie. You
reveal your magic cage. Inside, flying around, for all to see when they
peer through the window in the box, is a very angry faerie!

339
340 Conjuring with Computation

How It Works
This illusion is based on a Victorian stage illusion called Pepper’s
Ghost. In the original trick, a large glass sheet lay at a 45-degree angle
between the audience and the stage. Some actors acted just as normal
on the stage. Others lay on trolleys on the floor, out of sight, but on
the audience side of the glass. When powerful stage lights were shone
on them, they appeared (and disappeared) as ghostly figures on the
stage. This works because the light reflected on them bounces onto the
glass and is reflected out to the audience, making those actors appear
as though they are on the stage.
What is happening is the glass acts as a half mirror. Light from
behind it can pass through, but also light from below is reflected at 45
degrees out to the viewer too. Because our brains assume light travels
in a straight line, the reflected image appears to be in front of the back
wall. You can see a weaker version of the same effect at night just by
looking out the window. You see a reflection of the room superimposed
on the garden next to lit objects actually in the garden.

Trapping a faerie
To create your own (small) version of the illusion, you need to make
your own Pepper’s Ghost box: the faerie cage. For the faerie itself, you
need a smartphone with an animation of a faerie. It doesn’t have to be
a faerie, of course, you can “trap” any creature that you can find or
create an animation of. You could even film your brother or sister lit by
a lamp shining on them in a dark room (if you prefer, your story to be
about shrinking and trapping your irritating sibling). If you don’t have
an animation or film, then you can just use a still image on your phone
(and explain the faerie is trapped frozen in suspended animation). The
phone ensures you have a brightly lit image, equivalent to the bright
stage lights of the original effect, which is important if it is to work.

Making the faerie cage

Paint the inside of a small, strong cardboard box matt black. Ideally,
it should be just large enough to take a smartphone lying on the base.
 The Faerie Cage: Augmented Reality 341

Cut a large rectangle out of one side (the lid of the box), leaving a rim
all around the box. That is the window through which your audience
members can view your trapped faerie. Add bars if you want it to look
more like a cage.
Now cut a sheet of stiff Perspex (from a DIY or craft shop) to fit
diagonally inside the box at 45 degrees. The top of the Perspex should
be towards the audience, above the hole in the box, and the bottom of
the Perspex should be at the back of the box. It splits the box into two
diagonally. Your viewers can see the back wall of the box through the
window and through the transparent Perspex (Figure 48.1).
Now for the secret part. You place your smartphone (ideally, black
or covered so only the screen is visible) on the floor of the box, under the
Perspex. It should be playing the animation of the faerie on a loop. The
phone itself must be out of sight of the viewers who look at the back
wall of the box through the window. An overhanging lip at the bottom
of the window may help hide it from someone viewing from the front.
If you play an animation of an angry faerie on a black background,
then the light from the faerie will bounce off the Perspex and appear
to be flying, trapped, inside the cage. The image from the smartphone

Figure 48.1. A Pepper’s Ghost box. Because our brains assume that light travels in
a straight line, the image from the hidden camera that bounces off the half mirror
appears to come from the back of the box.
342 Conjuring with Computation

appears as if floating in the middle of the box. For a more complex

version, you could put props like a miniature bed behind the Perspex
to make it appear the faerie is in a prison room. Experiment with what
looks best.
TO PREPARE the Faerie Cage:
1. Create or find an animation of whatever you want to trap in your
cage such as an angry faerie.
2. Create a Pepper’s Ghost box and place a smartphone at the
bottom playing the animation on a loop.
3. Cover the box with a cloth.
TO DO the Faerie Cage:
1. Tell the audience you have trapped a faerie in a special magic box.
2. Whip off the cover to show the box.
3. Allow people to look at the faerie through the window, one person
at a time.

Thinking Computationally
Essentially the same illusion as Pepper’s Ghost is used for head-up
displays, where a virtual computer display is superimposed over the real
world. For example, my car’s Sat Nav projects arrow directions in this
way so they float out in front of the car as I drive. The arrows appear
at the right depth so I can focus on them while still paying attention
to the road. Fighter pilots have more complicated versions so they can
see information while still looking out the cockpit window. The virtual
band, Gorillaz, performed, apparently live on stage, interacting with real
singers like Madonna in the same way. You can also see the illusion in
museums when ghostly characters appear in dioramas and talk to you
about the past.
The same idea of combining the real world with the virtual world
forms the basis of augmented reality, though modern versions may
use different technology to achieve the same effect. Programmers not
only can now create whole virtual worlds, but they can also superimpose
a virtual world onto the real one using programs that merge live film
with the computer-generated imagery. Another way is to have smart
 The Faerie Cage: Augmented Reality 343

glasses that allow you to see the real world but where the programmers
superimpose imagery, like Pepper’s Ghost, onto the lenses so appearing
out in the world.

Interlude: John Henry Pepper

and Henry Dircks
When the very first Pepper’s Ghost appeared on stage on Christmas
Eve, 1862, it stunned the audience. A ghostly figure appeared with
the actors out of thin air, interacted with the other, more solid,
characters, and then disappeared in an instant. This was more
than just magic: it was miraculous. It was so amazing that some
spiritualists were convinced John Henry Pepper (1821–1900) had
discovered a way to summon spirits. This was no dark seance where
ghostly effects happen in a darkened room. Nor did it need the
virtual world of a computer (they didn’t exist yet!). It was on a
brightly lit stage in front of everyone’s eyes . . . Magicians are also
often scientists and engineers. They use new innovations, but also
invent some themselves. Often they patent their ideas rather than
keeping them secret. East Londoner and scientist Pepper co-invented
the effect with Henry Dircks (1806–1873), a Liverpudlian engineer.
They patented Pepper’s Ghost in 1863 as an effect to use in the
theatre. It was originally invented far earlier though; Giambattista
della Porta (1535–1615) described it in his 16th century Italian book,
Magia Naturalis.
This page intentionally left blank
 Chapter 49

The Down-Under Deal and

Robotic Reveal
Robots

Conjuring
A good reveal can make a trick and is also where you can have
some creative fun. Here, despite a volunteer having a free choice
of where to split the pack, the special Down-Under Deal leads to
a final selected card that no one could have predicted. It turns out
that the card was forced by the magical influence of your robot
assistant down under the table.

Computation
Robots have long been the subject of science fiction but are now
reality. Industrial robots work on production lines, building cars,
and anyone can buy a robot vacuum cleaner to clean their house
while they are out. Robotics researchers are building robots that
mimic all kinds of life and, using soft robotics; they are no longer
the hard metallic machines of science fiction.

345
346 Conjuring with Computation

The Trick
You explain that you are going to show a special magical deal called the
Down-Under Deal taught to you by an Australian magician (or European
one if you are from, for example, Australia or New Zealand). You shuffle
a pack of cards and then spread the cards in a line, face up on the table.
Point out that it is a normal pack of cards. Saying that the deal will
take too long with this many cards, ask a volunteer to point to a card
“somewhere in the middle” of the spread deck and then discard half the
pack beyond that position. Note that it was their free choice as to the
cards discarded as you do so.
Take the remaining cards and, forming them back into a pack,
explain you are now going to do the special Down-Under Deal that
will leave you with a single chosen card no one could have predicted.
With the pack face down in your hands, deal cards into two piles
saying “Down” as you put a card on the first, face-down, pile and
“Under” as you place the next on the second face-up pile. When all
cards are dealt, discard the face-down pile. Explain that the audience
should now be able to see why you wanted fewer cards as it is a very
slow deal. Deal out all the remaining cards from the “under” pile, into
two piles in the same way, chanting down-under-down-under . . . as you
place the cards, then again discard the down pile. Keep doing this until
only one face-up card is left in the un-discarded “under” pile.
Declare this to be the final chosen card that no one could have
predicted. Pick up the card and show everyone what it is. Point out
again that the volunteer had a free choice of where to split the pack
and turn over the pile of discarded cards, showing the top few cards to
the audience. Note that if the pack had been split one place further, it
would have been a different card chosen.
Finally, you explain that the deal is called the Down-Under Deal,
not just because of the way you do the deal, but because the final result
is influenced by whatever is down under the table, and throughout the
whole trick, your assistant has been down under the table, holding a
specific card to try and influence the final choice. You ask your assistant
to reveal that card . . . and at that moment, a robot pushes itself through
the table cloth draped over the table and lifts up a large card to
 The Down-Under Deal and Robotic Reveal: Robots 347

show everyone. It is the same as the final card left on the table. The
card chosen was magically determined by the influence of the card your
robot helper was holding under the table during the trick.

How It Works
Before the trick, place the card you intend to predict (for example, the
8 of Hearts) in the 16th position from the top of the pack. Place an
identical card in the grip of the robot and place it under the table. Have
a hidden button that activates the robot somewhere easy to press such
as at the back of the table.
The shuffle at the start is just a False Top Overhand Shuffle
(Chapter 18). This leaves the top 20 or so cards in the same position,
so leaving the 16th card in that position. If you are unsure of doing this
successfully, then you can miss out the initial shuffle.
The mathematics behind this trick is a simpler version of that used
in the Free My Three trick of Chapter 40. By discarding every second
card repeatedly, you will end up with a card that started in a specific
position. Starting with 52 cards, so positions 1–52 as below:
1, 2, 3, 4, . . . , 50, 51, 52.
we first split the pack somewhere in the middle. It is important that
the split happens somewhere between the 16th and 32nd cards, leaving
the 16th but discarding the 32nd. If the volunteer points somewhere
in the middle, then you will be in this range. We therefore end up, after
the discarding part of the pack, with cards such as:
1, 2, 3, 4, . . . , 14, 16, 18, . . . , 30, 31.
The precise last card is unknown, but the above is the worst case that
could happen. Note that the critical 16th card is still in the 16th position
from the top of the pack.
Deal the odd-position cards into one pile (face down) and the
even-position cards into another (face up). Discard the face-down pile
containing those in the odd positions. This leaves cards from the original
even positions only:
2, 4, 6, 8, . . . , 14, 16, 18, . . . , 30.
348 Conjuring with Computation

Repeat this to leave, after the second deal:

4, 8, . . . , 12, 16, 20, . . . , 28.
and then again leaving:
8, 16, 24.
The final deal takes out the odd-position cards of these (so first and
last), leaving just the planted 16th card:
16.
The final reveal can make or break a trick and is a way you can add
your own creativity and personal stamp on an existing trick. We have
played with different reveals for this trick in different shows and magic
workshops. My (Paul’s) favourite reveal was kindly filmed (for Peter) by
games designer and magician, Richard Garriott. The film shows Richard
finding the same card as the one we end up with on the table, buried in
the snow at the North Pole during a trip of his there.
In the 19th century, French clockmaker and magician Robert-Houdin
inspired this version of the reveal with his famous trick: the Marvellous
Orange Tree. It involves a wonderful reveal using an automaton, an
intricate mechanical device. Automata were commonly designed to
mimic life in some way. They date back to the Ancient Egyptians and
Greeks.
His Marvellous Orange Tree trick involved destroying a handkerchief,
only for it then to reappear unharmed. The reveal involved an automaton
of a tree that started bare but grew real oranges, the last one of which
when cut open contained the handkerchief (see a version of the trick in
the film The Illusionist). Robert-Houdin often incorporated automata in
his tricks, blurring science, illusion and technology. Was it really done by
a machine, or was it part illusion? Perhaps the most famous blurring of
this machine–illusion boundary was that of the 18th-century Mechanical
Turk of Wolfgang von Kempelen. It was presented as a machine that
could beat humans at chess, though in fact was a hoax: it was all
illusion. Of course, now computer science has overtaken the magic, and
chess computers are easily the best chess players on the planet. What
astounded 18th-century audiences, and could then only work by illusion,
is now just everyday.
 The Down-Under Deal and Robotic Reveal: Robots 349

The modern equivalent of using automata in tricks would be to use

robots as in this trick. Perhaps you can build a computer-controlled
robot with the same artistic flair as Robert-Houdin for an even more
amazing reveal, or invent tricks that use robots as an integral part of an
illusion.

The Magical Algorithm

TO PREPARE the Down-Under Deal and Robotic Reveal:
1. Place a specific card in the 16th position from the top of a shuffled
deck of cards.
2. Place a large version of the same card in the grip of a robot,
controlled by a hidden button at the back of the table.
3. Place the robot with the card under the table, hidden from the
audience by the table cloth.

TO DO the Down-Under Deal and Robotic Reveal:

1. DO the False Top Overhand Shuffle.
2. Spread the cards face up on the table.
3. Ask a volunteer to point to a card somewhere in the middle.
4. Split the pack at that point.
5. Discard the half from the bottom of the pack, forming the
remaining cards back into a deck.
6. DO the Down-Under Deal.
7. Show the final card to the audience.
8. Point out that the volunteer had a free choice, and you could
have ended up with any card.
9. Say that this is called the Down-Under Deal, not just because
of the down under pattern, but because things down under the
table can control what happens on the table. Your assistant who
has been under the table holding a controlling card throughout
the trick will now prove this.
10. Press the button so that the robot rolls forward from under the
table, holding up the matching card.
350 Conjuring with Computation

TO DO the Down-Under Deal:

1. DO THE FOLLOWING 4 TIMES:
a. Pick up the remaining cards.
b. DO THE FOLLOWING UNTIL all cards are dealt:
i. Deal a card face down, saying “Down”.
ii. Deal a card face up, saying “Under”.
c. Discard the face-down cards.

Thinking Computationally
Robert-Houdin’s automata that mimicked life were sometimes pre-
sented as being magical in their own right: how could a machine possibly
do what they did? Other automata were mixed with illusion in more
complex tricks. Given humanity’s long obsession with creating machines
that mimic life, it is natural, once computers existed, we would want to
give these machine brains bodies too.
The science-fiction versions of robots are often either clearly
machines or humanoids, sometimes both. Modern robotics is far more
varied. Real robots come in all shapes and sizes. While many of the
robots now used day to day are very much functional machines with
functional forms: working in factories manufacturing cars, in warehouses
collecting the items you have bought online, or scuttling around the floor,
vacuuming the carpets, many others mimic life, as did the automata.
Researchers have created cockroach-like and flea-inspired robots, fish-
shaped robots, snake-like ones, and robots modelled on octopuses, to
give just a few examples. Animals need to fit many life niches and
consequently have evolved elegant solutions to wide-ranging problems
that work far better than human-engineered solutions. This has led to
the growth of biologically inspired computing, including in the area of
robotics. Just as automata mimicked many different forms of life, robots
do too. If you want a robot that can travel over difficult terrain, then
building a variation of a car or a human may not be the best solution.
Animals travel across terrain they are adapted to in a multitude of more
flexible ways than our vehicles; think of goats on cliffs, for example,
or gibbons swinging through the forest. If you wish to move through
water, then there are fishes that are far more manoeuvrable than boats
 The Down-Under Deal and Robotic Reveal: Robots 351

and submarines. If you want a machine that can get in and out of
tight spaces, then just look at the octopus, which does so in ways that
would be a boon for many a magic trick if only a human assistant
could do it. Animals provide many wildly different solutions to problems
like locomotion, so learning from them can lead to novel engineered
solutions.
One area of biologically inspired robotics is soft robotics. This is
about making robots that are soft, not hard, fixed and rigid, and can
also change shape and size. As a result, they can be much more flexible
and safer around humans than stereotypical hard robots made of metal.
There are many sources of inspiration for soft robotics including the
octopus, spiders and plants too. Plants use changing pressures of liquids
in cells, induced, for example, by light levels, to change their form to fit
needs. It is a mechanism like this that allows them to follow the Sun.
Soft robots can use similar systems to move using soft materials.
It is not just the form of animals that researchers copy but also
their behaviour. For example, swarm computing aims to copy the way
birds fly in massive, swirling flocks each following simple rules that
maintain the flock. Flocks of drones might follow similar rules to help
maintain coverage of an area being monitored. A more futuristic idea is
for flocks of tiny, microscopic robots to be injected into the body and
work together in swarms to attack cancer cells.
Other researchers explore different uses of robots, creating actor
and stand-up comedian robots to explore the art of performance. Even
here, Robert-Houdin was there first, having created an automaton to do
the cups and balls magic trick. Robots with general-purpose computer
brains and fine dexterity as well as programmed showmanship offer the
possibility of flexible magician robots able to do a wide range of tricks.
A computer brain might even make tricks possible for a robot magician
that no human would be capable of. Showmanship and illusion would
be needed to make it magical though.
By copying and drawing inspiration from animals, computer sci-
entists have found innovative ways to solve problems. Whilst seeing
machines mimic life perhaps does not seem like magic itself anymore,
we still find such machines and their abilities intriguing. There is also
scope for mixing robots with illusion to make wonderful tricks if you have
352 Conjuring with Computation

the creativity to do so. Or perhaps rather than inventing robotic tricks,

your interests would lie in making a robot magician do the performances
for you.

PAUSE: Ismail al-Jazari

Ismail al-Jazari (c.1136–1206) was a Muslim scholar who lived in
Mesopotamia in the 12th century and is famous for his humanoid
automata, amongst many other things. They have led to him being
called the father of robotics because he was interested in building
such machines that had a practical use. The Ancient Greeks by
contrast were more interested in automata as illusions and for
entertainment. Al-Jazari’s inventions included a waitress “robot”
that served drinks and one intended to help with ritual handwashing.
 Chapter 50

The One True Sovereign

Curtain Call and Creativity

Conjuring
We close with one last trick. Are you the one true Sovereign of
the British Isles: can you remove Excalibur from the anvil?

Computation
We overview the main themes of the book, but also explore one
last link to computing: how can you be sure someone is who they
claim to be?

The Trick
Tell the story of Excalibur, the magical sword set in an anvil, on top of
a stone, part of the legend of King Arthur. Merlin predicted that only
the true sovereign of Britain, descended from Uther Pendragon, would
be able to pull the sword from the anvil. Merlin’s magical test showed
that the teenage Arthur was the rightful King.
Explain that you have found a new Excalibur, a sword in an anvil.
Pull back the curtains to reveal a sword protruding from an anvil,
complete with plaque stating that only the heir of Uther Pendragon,

353
354 Conjuring with Computation

and so the one true Sovereign of the British Isles will be able to remove
the sword from the anvil.
Allow anyone who wishes to show their strength to try to pull the
sword from the anvil. When they fail, apologise that they are clearly not
royalty. Once everyone who wants to try has failed, say that perhaps
you should try too. Without any effort at all, you pull the sword free.
Proclaim yourself heir to the throne. If anyone is sceptical, replace the
sword and let them try again.

The Magical Algorithm

1. Explain the story of the sword in the anvil and that you have found
Excalibur.
2. Reveal the newly discovered sword in the anvil.
3. Allow volunteers to attempt to pull Excalibur from the anvil (who
fail to do so).
4. Explain that this shows they are not royalty, not the true Sovereign
of the British Isles.
5. Take hold of Excalibur and pull it out of the anvil.
6. Look surprised and declare you must be the true King/Queen of
Britain.
7. Replace the sword and allow anyone who wishes to try to remove
it again.

How It Works
We will leave you to work out your own way to make this trick work
for now, though building it will take ingenuity, work and lots of both
engineering and computing skills.

Thinking Computationally
Algorithms
This book has drawn out the many links between conjuring and com-
puting. Self-working tricks are essentially algorithms for doing magic.
 The One True Sovereign: Curtain Call and Creativity 355

Both are sequences of steps that if followed precisely guarantee to

achieve a result. In fact, some useful computing algorithms are the
mechanism behind actual tricks, like error-checking codes and search
algorithms. Showmanship is just added to make the effect of the
algorithm seem magical. Both magic and algorithms are ultimately
underpinned by mathematics. Inventing tricks and inventing algorithms,
so writing programs, involve the same basic skills of algorithmic
thinking. It is the skill of being able to create those series of steps and to
be sure they always work as well as then write them out in a given precise
notation. This requires logical thinking and attention to detail. Both
magicians and programmers need to be sure the steps do always work,
and that requires evaluation skills. In both cases, testing (just trying
it out some number of times) is complemented by rigorous argument
and proof. These are the basic skills of computational thinking.

Decomposition, abstraction and generalisation

Just as programmers use more advanced computational thinking skills,
so too do magicians. Programmers decompose programs they write
into named, reusable and more general parts. They create generalised
versions of parts that can be used in many situations, and build larger
programs from smaller, named parts they already know work.
Similarly, magicians name simple effects that they use in tricks and
build tricks by combining together different existing components to make
more powerful effects. To create their own versions of a trick, they strip
it down to its basic core mechanism and build it back up in new ways to
make different tricks. Both programmers and magicians therefore rely
on decomposition, abstraction and generalisation.

Data representation
Another important aspect of developing software is the appropriate
choice of data representation: how is information organised? It needs
to be done in a way that facilitates the task. This is also important
in many tricks, with different representations used as the basis for
organising information in tricks. Whether it is in the organisation of
356 Conjuring with Computation

a pack of cards in a certain order needed for a trick to work, in the

way information is coded in the props used, or in the way information
is passed between magician and assistants, the way information is
represented is central to many tricks.
In programming, key things about the choice of representation
include how easy the operations are to do, how quickly they can be
done, and how efficient their use of storage space is. In a magic trick, the
issues are similar. You want a representation that makes the trick easy
to perform perhaps or at least possible given constraints on the time and
abilities of the magician to do it flawlessly. However, another important
consideration is that the representation chosen makes it difficult for the
audience to notice vital information about how the trick works.

Computation and cognition

One view of cognition, our ability to think, is that it is just computation.
Our brains are biological computation machines. Computation is
about the manipulation of symbols that we then give meaning to.
Symbols can be all sorts of things, including physical objects like cards
in a deck, that we give meaning to. Symbols can exist in a computer
or out in the world. This is the main underlying link between conjuring
and computation, they are both manipulating symbols, so conjuring is
just computation with the aim of providing surprising and entertaining
effects.
This ties in with Edwin Hutchins’ ideas of “cognition in the wild”. We
extend our cognitive abilities by creating a physical system of symbols
in the world. We use cultural artefacts as ways to extend our cognitive
abilities such as our memories, and to translate information into new
representations that make the tasks we do easier. Prayer beads help
us count, a shopping list helps us remember, an abacus helps us do
arithmetic, and a list of turns helps us navigate to a destination. In
this view of cognition and computation, conjuring is using cultural and
information artefacts, symbols in the world, to make it harder rather than
easier to think: harder to realise what manipulations of the symbols are
really going on, whether information in people’s heads or in the world
as in cards, coins or other objects.
 The One True Sovereign: Curtain Call and Creativity 357

Psychology and human–computer interaction

Tricks are more than the steps. They involve engineering a system where
everyone in the audience makes the same mistake at the same time.
This relies on a deep understanding of both cognitive psychology
and social psychology: of the effects of attentional resources, memory
limitations, social contexts, and so on. Magicians go out of their way to
ensure errors made by the audience are undetected until the final reveal.
Creating usable and used programs involves a similar understanding of
psychology, though to use it in a way that ensures no one using the
software makes mistakes, or if they do, then they immediately see and
correct the error. This is done using similar techniques to the magicians,
just with the opposite aim: for example, controlling attention so that
everyone looks at the right place at the right time rather than looking
at the wrong place.
This leads to the direct link between conjuring and inter-
action design and the study of human–computer interaction.
Human–computer interaction applies psychology to formulate design
principles for how to make computer systems usable, based on models
of how we as users of software get things done. We work from goals,
creating plans of action that we then try to follow to do tasks based
on signifiers from an interface. We monitor whether things are going
to plan based on feedback given by the system. This leads to design
principles such as using and communicating clear conceptual models
of how systems work, ensuring the underlying system state is visible so
people can track their progress, giving clear feedback, being consistent,
and so on.
Magicians just turn these design principles on their heads. Instead of
avoiding confusion, they sow the seeds of it, ensuring the audience lose
track of what has happened and is happening: using conceptual models
to confuse, not to enlighten, ensuring critical aspects of the system state
are hidden so tracking what is happening is impossible, giving misleading
feedback, doing things differently to the way expected, and so on.
Engineering of user experience is similar in both conjuring and
computing, too. When using computer software, delight occurs when
things are surprisingly simple. In a magic trick, delight comes from
surprises too, but because the person lost track so that the reveal is
358 Conjuring with Computation

surprising! Good system design, whether of computing or conjuring,

involves engineering-appropriate experiences.

Computational agents versus practice

There are other differences beyond the opposite ways psychology is used,
between writing programs and inventing tricks. Computer scientists have
a notion of a computational agent: something that follows instructions
perfectly and blindly, with no recourse to thinking about what it is doing.
Computational agents get things done totally by unthinkingly following
the algorithm with no improvisation. Computers are built to be such
computational agents.
Magicians, when performing tricks, act as computational agents in
that they also just follow the steps. Once well-practised, they should need
no improvisation. All possibilities should be covered. A novice can follow
the steps of a self-working trick, without thought and without knowing
how the magic works, and achieve the effect. However, magicians do
need to practise lots and lots to be able to follow the steps of a trick
as perfectly as possible. Some tricks, even though just following steps,
require a great deal of practice to master the levels of dexterity and
timing needed, whether it is palming a coin or doing a perfect riffle
shuffle. Computers of course do not need to practise to execute software,
they just follow the instructions.

Precise languages
Unlike magicians, computers follow instructions in a programming lan-
guage, a precise mathematical language. A large part of programming
is in writing the steps in the constructs that are available in the language
being used.
Magicians do not rely on special languages the same way, so
the important programming step of writing the instructions in such a
mathematically precise language is omitted. Humans work out tricks
based on natural language explanations and fill in the gaps themselves if
necessary. If a robot were to become a magician, then the tricks would
need to be set out precisely in an appropriately precise language. An
 The One True Sovereign: Curtain Call and Creativity 359

alternative is a machine would need a similar level of intelligence to

humans, but it would still ultimately be following instructions given in a
precise language, just ones for intelligence not for tricks.

Scale
There is also a massive difference in scale between tricks and programs.
Programs can be millions of lines of code, millions of steps long, whereas
even the most complex of magic tricks is likely to be only tens or
hundreds of steps; after all, you do not want to bore your audience,
while they wait for the pay-off of the trick! Computational thinking
techniques that are used in simple ways with tricks are absolutely vital
to manage the complexity of large programs.

Showmanship
Showmanship is all-important in magic, but less so in computing. It is
the ability to engage and entertain the audience, to draw them into the
world you are creating. It is more than just following the core algorithm
of a trick. It is about the presentation around that: the way a trick is
weaved into a story, and then the way that story is told, the timing, the
flair. Showmanship makes the performance of a trick extra special.
The equivalent to this in computing is user experience (UX)
design. Something akin to showmanship is just one aspect of this.
Similar techniques may draw people into software that aims to entertain
like games. Those using the software expect to have an enjoyable,
entertaining experience in a similar way as a magic trick. Educational
software is often gamified, for this reason: trying to make it entertaining
so children will want to use it repeatedly, rather than it being a chore.
Just as magicians want to engineer delight, often software engineers do
too. After all, if a shopping website or app is a delight to use, you are
more likely to return and so buy more.
However, user experience design is about more than this kind of
showmanship. It depends on the specific task at hand. For some software,
a good experience is about being able to do the job quickly and easily
with the minimum of fuss; about the software being very easy to use,
without frustration. A good experience may even be an experience where
360 Conjuring with Computation

you do not notice you have interacted with software at all. For example,
to open my car, I just grip the door handle, and it recognises me and
unlocks automatically. That is far better than having to dig around in
my pockets for keys.

Security and privacy

There are big overlaps between cyber security and a variety of magic
tricks. One of the aims of cyber security is to prevent leakage of
information, allowing only certain people access to resources, and often
ensuring that no one else even knows they exist. Many mentalism
tricks rely on secret codes to pass information between magician and
assistant, whether by known legitimate channels or hidden covert
channels. Steganography techniques hide the fact that any information
is passed between magicians.
Some tricks actively involve a magician gathering information about
members of the audience. In others, they just pretend to know things.
Information about people can now be gleaned from a variety of sources,
and big data techniques mean your privacy is lost if enough people,
even ones you have never met give up information about themselves.
Undoubtedly, such technology can form the basis of new tricks too.

Science and technology

Magicians have always been at the forefront of science and technology:
having a better understanding of physics, mechanical engineering or
electronics than their audience can be a basis of a trick if the effect
is combined with illusion. This was part of what drove the golden
age of magic and illusion, from Victorian times into the early 20th
century. Pepper’s Ghost, for example, uses some simple physics of light;
automata-based tricks involved intricate mechanical engineering, and
there are many opportunities to base tricks around electronic gadgetry.
Computing gadgetry offers yet more new opportunities.
Sometimes magicians are ahead of the game, rather than just
following science and engineering. For example, a hot research topic
at the moment is wearable computing, where electronic gadgetry is
 The One True Sovereign: Curtain Call and Creativity 361

built into clothes, but magicians have been doing this for years. What
tricks can you imagine doing with electronics, whether hidden controls,
sensors or displays built into your stage clothes? What might you do with
body-centric networks that use a person’s body as a network to pass
signals between different gadgets detecting things you touch? How about
buttons for controlling things built into your skin, or brain–computer
interfaces that can pick up a person’s simple thoughts and pass them
to a computer. A magician who is ahead of the game with mathematics,
science and technology can always invent new tricks and new ways of
doing old ones that baffle and entertain.

Creativity
Being a magician, inventing new tricks and adding new twists to old
ones, is not just about being on top of technology but all about
creativity too. Computer science is a similarly creative subject. In fact,
some computer scientists have gone a step further and see creativity as
something to investigate algorithmically. Machine creativity, computer
programs creating poems, jokes, songs, stories, films, and so on, is a
major area of research.
Machines have long ago been created to perform tricks but in future
they could create new ones too. Researchers have started to explore this
with artificial intelligence programs that develop the most magical
version of a trick. PhD student Howard Williams developed a program
modelled on evolution that created its own version of a jigsaw trick: the
teleporting robot trick (Chapter 11). It used a kind of algorithm called
a genetic algorithm to “breed” better and better versions of the trick,
selecting for ones that would appear most magical, exploring many more
possibilities than any human magician trying to develop a trick could.
True computer magicians are still in the future, but one day they may
be inventing their own completely novel tricks.

An authentication problem to finish

If you want to become a magician, inventing new effects and mecha-
nisms, then you would do worse than to follow science, computer science,
362 Conjuring with Computation

electronic engineering and technology closely. Hone your skills at making

gadgets, as many magicians over the years have. This approach was
exemplified by Robert-Houdin, but to make a trick magical, you will
also need to be able to combine creativity with the basic effects: an
ability he had in spades.
One way to do Excalibur magic, for example, is based on that of
the original version: Robert-Houdin’s “The Light and Heavy Chest”. He
created a magic box that he explained was a perfect way to prevent theft
as it sapped the energy of even the strongest people who attempted to
open it. An innocent child, however, would not be affected and could
open it at will. The strongest volunteers from the audience would try,
but no one could lift the box. Once all had given up, a child volunteer
would then lift it with ease. How could this be done? Electromagnetism
had recently been discovered. If you pass an electric current through a
coil of wire, it creates a magnetic field. Wrap the coil of wire around
an iron bar and the bar becomes a strong magnet. The more times the
wire is wound around it, the stronger the magnet you make. However,
switch off the electricity and the magnet switches off too. All it took
was a flick of a switch to turn on and off powerful electromagnets that
held the iron box to the table.
A modern version could add to the trick, and undoubtedly Robert-
Houdin would have done so, had he been alive in the 21st century. In
computer science terms, Excalibur’s role in the myth was to determine
the identity of the person holding it: were they the true Sovereign of
Britain or not? If they were, then and only then were they allowed to
remove it from the anvil. Excalibur was just an authentication device.
Today, it could be done without the need for a secret on–off switch
and work actually using modern authentication techniques. A wide range
of authentication devices have been invented to do this, so take your
pick! Keys and passwords do not really work in this situation, but keyless
locks, for example, as found now on many cars allow you to open a door
just by gripping the door handle, as long as you have an electronic key
in you pocket. Sensors in the handle trigger the car’s computer to check
if the key is close enough to unlock. The hidden key broadcasts a code
(the equivalent of a password) by radio. If the right code is detected,
then the door is unlocked.
 The One True Sovereign: Curtain Call and Creativity 363

An alternative approach is biometrics. Biometric authentication

uses unique features of people like their fingerprints to check who they
are and what rights they have. My laptop has such a finger pad which
I can just touch to unlock the laptop avoiding typing a password.
My tablet on the other hand uses facial recognition. It allows me
in because it recognises unique features of my face, just as some fast-
track passport systems at airports do to get people through immigration
quickly. Yet another alternative might be to use voice authentication,
where unique features of your voice identify you. (That would also be
useful to make magic words actually work!)
Any of these (and more) could be used as a way to allow a trick to
work for you and only you, as with Excalibur. However, once technology
is widespread, it is not as simple as just building it into a trick.
Showmanship is needed to distract from what is actually happening
so that it still seems magical. Part of this is to embed the trick in a
story (such as searching for the rightful King or Queen of Britain). Part
is by combining it with other illusions so that an apparent obvious way
appears to be impossible. That is where creativity comes in. Over to you.

Interlude: Richard Garriott

Richard Garriott made his fortune as a programmer and games
designer, creating massively multiplayer online role-playing games
(a term he invented). He is also a keen magician and astronaut,
travelling to the International Space Station in 2008, and while
there, he performed the first-ever low-gravity magic show in space,
broadcast back to earth. There are always new frontiers for science
and engineering, and with them, new frontiers for magic.

Invent your own ways to present tricks, invent new versions

from old and invent completely new tricks based on your knowl-
edge of and skills in maths, science, technology, and especially
computing.
This page intentionally left blank
 Further Reading

Find resources to accompany the tricks in this book at

https://conjuringwithcomputation.wordpress.com
There is lots more of our magic and computer science at
http://www.cs4fn.org/magic/
Magical computing resources for teachers can be found at
https://teachinglondoncomputing.org/magic/

The History of Magic and Magicians

Hiding the Elephant, Jim Steinmeyer, Arrow Books, 2005.
And that’s renaissance magic . . . , Lucy McDonald, The Guardian,
https://www.theguardian.com/world/2007/apr/10/italy.books, 10 Apr
2007.
Magic: 1400s–1950s, Noel Daniel (Ed.), Taschen.
Famous Female Magicians Through the Ages, Katherine Mills,
https://www.katherinemills.co.uk/blog/most-famous-female-
magicians-of-all-time
David Copperfield’s History of Magic, David Copperfield, Richard
Wiseman and David Britland, Simon and Shuster, 2021.
The Discoverie of Witchcraft, Reginald Scott, 1886. Available from
https://archive.org/details/discoverieofwitc00scot/

365
366 Conjuring with Computation

Dot-dash-diss: The gentleman hacker’s 1903 lulz, Paul Marks,

New Scientist issue 2844, 2011.
https://www.newscientist.com/article/mg21228440-700-dot-dash-diss-
the-gentleman-hackers-1903-lulz/
Wikipedia. www.wikipedia.com

Self-working Magic Tricks

Self-working Card Tricks, Karl Fulves, Dover, 1976.
Self-working Mental Magic, Karl Fulves, Dover, 1979.
Self-working Table Magic, Karl Fulves, Dover, 1981.
Self-working Number Magic, Karl Fulves, Dover, 1983.

Magic Tricks in General

Magic and Showmanship, a Handbook for Conjurers, Henning Nelms,
Dover, 1969.
Martin Gardner’s Table Magic, Martin Gardner, Dover, 1998.
Encyclopedia of Impromptu Card Forces, 2nd Edition, Lewis Jones, 2004.

Optical Illusions and Op Art

Japanese Optical and Geometrical Art, Hajime Ōuchi, Dover Pictorial
Archive, 2000 (original edition 1977).
Bridget Riley, accompaniment to exhibitions of Riley’s work, National
Galleries of Scotland, Hayward Gallery, 2019.

Magic and Mathematics

Mathematics, Magic and Mystery, Martin Gardner, Dover, 1956.
Mathematical Magic, William Simon, Dover, 1964.
Magical Mathematics, Persi Diaconis and Ron Graham, Princeton
University Press, 2012.
 Further Reading 367

Magic and Computer Science

The Magic of Computer Science, Paul Curzon and Peter W.
McOwan, Queen Mary University of London, 2008.
The Magic of Computer Science II, Now We Have Your Attention, Paul
Curzon, Peter W. McOwan and Jonathan Black, Queen Mary University
of London, 2009.
The Magic of Computer Science III, Magic Meets Mistakes, Machines
and Medicine, Paul Curzon and Peter W. McOwan, Queen Mary
University of London, 2015.

Unplugged Computer Science

CS Unplugged. Tim Bell, Ian Witten and Michael Fellows (founders).
https://www.csunplugged.org/
Teaching London Computing, Paul Curzon, Peter W. McOwan and
William Marsh (founders). https://teachinglondoncomputing.org/

Computer Science
The Power of Computational Thinking: Games, Magic and Puzzles to
Help You Become a Computational Thinker, Paul Curzon and Peter W.
McOwan, World Scientific, 2017.

Human–Computer Interaction, Cognition

and Computation
Cognition in the Wild, Edwin Hutchins, MIT Press, 1995.
The Design of Everyday Things, Donald Norman, MIT Press, 1998.

Movies about Magic

The Prestige, directed by Christopher Nolan, 2006.
The Illusionist, directed by Neil Burger, 2006.
Now You See Me, directed by Louis Leterrier, 2013.
368 Conjuring with Computation

Novels about Magic

The Prestige, Christopher Priest, Simon and Schuster, 1995.
Carter Beats the Devil, Glen David Gold, Sceptre, 2001.
 Acknowledgements

Thanks to my family, for putting up with me constantly disappearing

to write and especially Daniel for his enthusiasm for my magic. Sam
Stein gave advice on the early chapters. Thanks also to all at World
Scientific, especially Rochelle Kronzek for seeking us out. Jonathan
Black, Jane Waite and Jo Brodie have worked on the cs4fn project
as we originally developed the magic ideas. I would also like to thank all
those who have supported the cs4fn project at Queen Mary University of
London and elsewhere, since its inception, and especially the Engineering
and Physical Sciences Research Council (EPSRC) who provided major
funding for cs4fn. We also developed many of the ideas of linking magic
and interaction design as part of the EPSRC-funded CHI+MED research
project on safer medical device design. Other major funders of our public
engagement work have included Google and the Mayor of London.
The Cyclops Eye was created by Kelly Burrows based on a design by
Peter, and the original illusion was made freely available for designers
to use by Hajime Ōuchi. Kelly was also the person behind the design
of our cs4fn magazine. Paul drew the teleporting robot. Wikipedia has
been invaluable in researching the book and particularly for learning
more about, and fact-checking, the background of the scientists and
magicians. Tim Bell pointed out some of my mistakes. Peter provided the
majority of the tricks from his lifetime studying magic. I am also indebted
to Jane McOwan and the rest of Peter’s family for their enthusiasm and
support for me to finish this book and for passing on his collection of
magic books; some of my favourites and those I have found most useful
learning about magic and its history are included below along with ones
I am aware of as being favourites of Peter’s. Peter would undoubtedly
have included others that I have missed, so my apologies for any such
omissions.

369
This page intentionally left blank
 Index of Computing Terms

A 205–206, 213, 216–217, 222,

A320, 218 232–233, 240, 248, 258, 269–270,
abacus, 12, 356 277–278, 287, 298–300, 306,
abstraction, 88–109, 117–120, 129, 313–314, 320–321, 327–328, 333,
134, 163, 255–265, 288, 291, 355 349–350, 354
abstraction, data, 163 algorithm, search (see search), 169, 355
abstraction, layers of, 117–120 algorithm, sort, 169, 283
abstraction, procedural, 95–109, 118, algorithmic thinker, 24, 28, 73
129, 134 algorithmic thinking, 5, 12, 46, 355
actual parameter (see parameter, alternating current (AC), 115
actual), 124, 151 Analytical Engine, The, 180
airline disaster (see disaster, airline), app, 359
218 application programming interface
alarm clock, 207–208 (API) specification, 134
algebra, 255–264 argument (see parameter), 121–125,
algebra, boolean (see boolean), 24 150–151, 153–154, 158
algebraic proof (see proof, algebraic), argument (see proof), 24, 53, 59, 63,
255–264 91, 153, 158, 255, 355
algorithm, 5–31, 34–39, 45–46, 51, argument, convincing (see proof), 53,
53–68, 83–84, 98–99, 102, 108–109, 59, 63, 91, 153, 158
112–114, 122–123, 125, 129, 134, argument, logical (see proof), 156, 158
140–141, 151, 153–161, 169, 175, argument, reasoned (see proof), 24
179, 183, 186, 206, 209, 247–253, argument, rigorous (see proof), 24,
263, 267, 272, 275, 280–282, 285, 53–68, 91, 153–161, 255, 355
292, 307–308, 313–315, 320–321, Ariane 5, 61
329, 354–355, 358–359, 361 arithmetic, 12, 24, 125, 174, 186, 356
algorithm, genetic, 161 artificial intelligence, 4, 148, 214, 329,
algorithm, magical, 9, 14, 21–22, 27, 361
32, 44, 55, 65, 72, 75, 81, 92–93, assertion, 114, 154–161, 249–252, 273
97–98, 102, 108, 112, 118, 122–123, assumption, 73, 111, 113, 222
128–129, 132, 139, 144, 146, 150, attention, 36, 84, 201–209, 211, 214,
158, 167–168, 172–173, 178–179, 232–233, 342, 357
183–184, 188–189, 194–195, attention to detail, 34, 45, 50–51, 355

371
372 Conjuring with Computation

augmented reality, 339–343 C

authentication, 361–363
Caesar cipher (see cipher, Caesar), 309
authentication, biometric, 363
calculator, 31–37, 125, 177, 179, 208
authentication, facial recognition, 363
call, function, 149–151
authentication, fingerprint, 363
call, procedure, 109, 134, 154, 156
authentication, voice, 363
card payment, 15
automata, 3, 348–350, 352, 360
cashpoint, 234
automata, humanoid, 350, 352
chaining assertions (see chain of
automated program verification (see
reasoning), 156–157
verification, program), 161
automation surprises, 215, 218 chain of reasoning, 156–157, 249–253
chess computer, 348
B chip-and-pin card, 15
chip-and-pin card machine, 15
big data, 325–331, 360
cipher, 4, 142, 303–309
binary, 181–187, 189, 195–197,
cipher, Caesar, 309
300–302
clarity, 94
binary arithmetic (see binary), 186
cockpit design, 218, 339, 342
binary number (see binary), 184–186
cockpit, fighter aircraft, 339, 342
biological computation machine, 356
code (see data representation),
biologically-inspired computing, 214,
165–169, 191–197, 303–310, 317,
350
355
biologically-inspired robotics (see
code (see program), 22, 37, 51, 61–63,
robotics, biologically inspired), 351
101, 105, 109, 114, 117, 121,
biometrics (see authentication,
124–125, 127, 141–142, 147, 253,
biometric), 363
272, 317, 322, 356, 359
bit (see binary), 196–197, 301
code, binary, 301
bit, parity, 196–197
bitmap graphics (see graphics, bitmap), code, error-checking, 195, 355
189 code, error-correcting, 191–197, 355
black hat hacker (see hacker), 331 code, reliable, 117
Bletchley Park, 142 code, safety critical, (see software,
body-centric network (see network, safety critical), 141, 255
body-centric), 361 code, secret, 303–310, 317, 355, 363
body, of a loop, 28–29, 249 code, spaghetti, 62
boolean, 22, 24 code walkthrough, 125
boolean algebra (see algebra, Boolean), code word, 169
24 cognitive psychology, 69–85, 199–244,
boolean expression, 22 356–358
boolean logic (see boolean), 24 cognitive science, 84, 207
boolean value (see boolean), 24 Colossus, 142
brain-computer interface, 361 comment, 111, 141
Bridges of Königsberg, 291–293 communication protocol (see protocol,
bug, 37, 41, 48–50, 61–62, 98, 125, 161 communication), 314
button, 22, 51, 79, 84, 208, 214, 225, communications technology, 174
243, 347, 349, 361 compiler, 37
 Index of Computing Terms 373

computation, 1–2, 4, 11, 16, 28, 36, data representation, 163–199, 285–294,
148, 169, 226, 253, 258, 281, 356 297–317, 355
computational agent, 11–12, 358 data representation, choice of,
computational thinking, 1, 44, 99, 104, 163–199, 285–294, 297–317, 355
119, 124, 165, 168, 265, 331, 355, data, store, 163, 180, 186, 189, 191,
359 195, 273, 285, 300–301, 308, 314,
computer, 1, 4–5, 11, 15, 24, 33–34, 323, 329
37, 49, 51, 77, 98, 125, 134, 142, data structure, 131, 143, 147, 151,
148, 149, 169, 174, 179–186, 189, 285–293
191, 207–208, 226, 235, 264, 273, data structure, graph, 285–293
281, 303, 314, 320, 322, 335, 343, data structure, list, 109, 114, 134, 147,
348–351, 356, 358, 361–362 169, 189, 272, 356
computer game (see game, computer), data, transmit, 191, 195, 197, 308,
50, 61, 134, 339, 348, 359, 363 317, 323
computer memory (see memory, debug, 48–49
computer), 62, 130, 179, 273 decimal number (see number system,
computer science, 1–2, 28, 30, 62, 84, decimal), 12, 175–181, 184–186
104, 161, 197, 214, 273, 291, 330, decision instruction (see selection),
337, 348, 361–362 19–25, 28, 62–63, 65–67, 138–139,
computer system, 2, 199, 208, 215, 184, 194, 333
218–219, 221, 234, 264, 295, 311, decode, 306–308
315, 357 decomposition, 63, 87, 95–101,
conceptual model, 221–227, 357 107–110, 119, 121–125, 134,
consistency, 357 143–160, 265, 275–283, 355
control structure, 15, 24–25, 28, 62 decomposition, parameterised,
control unit, 273 121–125, 150, 153
correctness, program, 15, 39, 41–51, delight, engineering, 99, 217, 237–244,
53–62, 111, 247–253, 255–264 357, 359
counter, 28–29 design metaphor, 132, 221, 225–227
counter-controlled loop, 28–29 design model, 224, 226
covert channel, 295, 316, 319–323, 360 design principle, 79, 84, 215, 217, 219,
creativity, 3, 68, 92, 115, 120, 125, 144, 226, 357
146, 345, 348, 352–353, 361–363 Difference Engine, 180
creativity, machine, 361 digit keypad, 208
critical system, 51, 62, 114, 141, 161, digital computer, 134, 142, 181
255 digital data, 197
cryptography, 303–309 digital image, 301
cursor, 208 digital money, 303
cursor-based number entry, 208 digital object, 187
cut and paste, 235 digital person (avatar), 339
digital world, 187, 189
D direct current (DC), 115
data abstraction (see abstraction, disaster, airline, 218
data), 163 disaster, rocket, 5, 61
data, preprocessing, 143, 147 divide and conquer, 275–283
374 Conjuring with Computation

drag and drop, 93, 235 formal parameter (see parameter,

drop-down menu, (see menu, formal), 124
pop-up/drop-down), 24, 79, 235 frequency hopping, 323
dynamo, 100, 294 function, 143–152
function call (see call, function/
E procedure), 149–151
easy to use, 69–85, 199–244, 357–359 functional programming language (see
edge (see graph), 289–293 programming language, functional),
148
EDSAC, 134
electricity, 3, 100, 115, 294, 362
G
electromagnetism, 3, 100, 115, 362
electronic telephone exchanges, 142 game, computer, 50, 61, 134, 339, 348,
encode, 166, 306–308 359, 363
end-to-end testing (see testing, game, massively multiplayer online
end-to-end), 51 role-playing, 363
engineering delight, (see delight, gamify, 359
engineering), 99, 217, 237–244, 357, generalisation, 121–125, 153–154,
359 265–273, 288, 291, 355
ENIAC, 273 genetic algorithm, 361
Enigma, 142 goal, 357
error-checking code, 195, 355 goto statement (see jump statement),
error-correcting code, 191–197 62
ethics (see professional ethics), 295, graph, 285–293
331–335 graphical user interface (GUI) (see user
evaluation, 39–68, 245–264, 355 interface, graphical), 73, 77, 93–94,
evaluation method, 51, 219 235
evaluation method, usability, 219 graphics, 189
execution, 25, 155, 248, 273, 322, graphics, bitmap, 189
358 graphics, raster, 189
eye-tracking technology, 209 graphics, vector, 189

F H
facial recognition, 363 hacker, 30, 114, 295, 311, 319, 322,
false value (see boolean value), 22–24, 328, 331, 335
28 hacker, black hat, 331
feature bloat, 79, 84 hacker, white hat, 331, 335
feedback, 217, 226, 357 hand dryer, 225
fingerprint recognition, 363 headlights, 233
flow of control, 15–16, 23–24, head-up display, 339, 342
28–29 high-level language (see programming
flowchart, 16, 23, 161 language, high-level), 37
focus of attention, 201–209, 211, 214 Hindu-Arabic numeral system (see
font, 94, 302 number system, Hindu-Arabic), 12,
formal logic (see logic, formal), 186 175, 179–180
 Index of Computing Terms 375

human-computer interaction (HCI), K

69–85, 93–94, 199–243, 357, 361 keep it simple stupid (KISS) principle,
human-computer interface (see 79–85
interface, human-computer), 69–85, keyless lock, 362
93-94, 199–243, 357, 361
humanoid automata (see automata, L
humanoid), 350, 352
layers of abstraction (see abstraction,
I layers of), 117–120
leak information, 295, 311–322
icon, 93–94, 226, 235 legitimate channel, 295, 311–317, 321,
If statement, 19–24, 65–67 360
include a library, 134 library, 109, 129, 131–134, 140
inductive reasoning, 247–253 limitations of brain, 69–85, 173,
infusion pump, 207–208 201–235, 357
instruction, low-level, 37 link, to click, 22
instruction, primitive 118–119 Lisp, 148
instructions (see algorithm, function, list data structure (see data structure,
procedure), 1–37, 50, 63, 93, list), 109, 114, 134, 147, 169, 189,
95–109, 117–135, 143–161, 163, 272, 356
248–249, 273, 358–359 logic, 24, 57, 66, 141, 148, 161, 186
interaction design (see logic, boolean (see boolean), 24
human-computer interaction), 69–85, logic circuit, 24
93–94, 199–243, 357, 361 logic, formal, 186
interface, application programming logic, programming, 161
(API), 134 logic-based specification language, 141
interface, human-computer, 69–85, logical thinking/reasoning, 39–68, 73,
93–94, 199–243, 357, 361 125, 156, 158, 186, 245–265,
interface, graphical user (GUI), 73, 77, 285–294, 355
93–94, 235 loom, Jacquard (see Jacquard loom),
interface, multimodal, 75, 77 180
interface, of procedure/programming, loop invariant (see invariant, loop),
71, 134, 137–141 247–253
internal state, 215–219, 224 low-level instruction (see instruction,
Internet, 2, 147, 189, 197, 294, 309, low-level), 37
320 low-level language (see programming
interruption, 234, 312 language, low-level), 37
invariant, loop, 247–253
M
J
machine learning, 329
Jacquard loom (see loom, Jacquard), maintainable software, 37, 103, 130
180 manual, 34, 134
Java, 11 Mariner 1, 61
jump statement (see goto statement), massively multiplayer online
62 role-playing game, 363
376 Conjuring with Computation

mathematical model (see model, number system, Hindu-Arabic, 12, 175,

mathematical), 255–263 179–180
mathematical puzzle, 85, 291 number system, place-based, 12,
meaning (of Barnum statement), 327 175–186
meaning (of data), 195, 301, 356 number system, place-value, 12,
meaning (of icons), 94 175–186
meaning (of program constructs), number system, positional, 12, 175–186
34–36, 114
meaningful part, 95, 98–99, 107 O
medical error, 207, 265 object, 189, 235
memorable, 94 object-oriented programming, 235
memory, computer, 62, 130, 179, 273 object-oriented programming language
memory, human, 229–235, 357 (see programming language,
memory limitation, 229–237, 357 object-oriented), 235
mentally tracking state, 64, 166, operating system, secure, 314
217–218, 357 operation, and, 24, 114, 154–155,
menu, pop-up/drop-down, 24, 79, 235 158–159
method evaluation, 51, 219 operation, not, 24
method, in programming, 147 operation, or, 24
method, scientific, 170, 302
method, usability evaluation, 219 P
Möbius strip, 13–17
parameter, 121–125, 150–151,
model, conceptual, 221–227, 357
153–154, 158
model, design, 224, 226
parameter, actual, 124, 151
model, mathematical, 255–263
parameter, formal, 124
model, user, 224–225
parameterised decomposition (see
multimodal interface (see interface,
decomposition, parameterised),
multimodal), 75, 77
121–125, 150, 153
multimodal software, 75, 77
parity bit (see bit, parity), 196–197
password, 114, 264, 309, 328, 362–363
N
past experience, 224
Navajo, 308 Pepper’s Ghost, 339–343, 360
network, 189, 191, 195, 288, 316, 329, permutation, 56
361 phone crash, 61–62
network, body-centric, 361 pixel, 74, 94, 189, 301
network, social, 288, 329 pixel art, 94
node, 289–293 place-based number system (see
notation, 31–37, 124, 141, 355 number system, place-based), 12,
number entry, 208 175–186
number system, 12, 175–186 place-value number system (see
number system, binary (see binary), number system, place-value), 12,
181–187, 189, 195–197, 300–302 175–186
number system, decimal, 12, 175–181, pop-up menu (see menu,
184–186 pop-up/drop-down), 24, 79, 235
 Index of Computing Terms 377

positional number system (see number programming logic (see logic,

system, positional), 12, 175–186 programming), 161
postcondition, 113–114, 153, 156–158, proof, 49, 53–68, 247–264
247, 249 proof, algebraic, 255–264
power transmission, 115 proof by cases, 63–68
precondition, 113–114, 123, 141, proof by induction, 247–253
153–158, 247 prospective memory error, 233–234
preprocessing data (see data, protocol, 114, 314
preprocessing), 143, 147 protocol, communication, 314
primitive instruction, 118–119 protocol, security, 114
pseudocode, 37
priority alarm display, 51
punch card, 180, 186, 281
privacy, 295, 322, 325–331, 360
Python, 11
procedural abstraction (see abstraction,
procedural), 95–109, 118, 129, 134 Q
procedure, 95–143, 146–147, 149, 151,
153–160 Quipu, 168
procedure call (see call,
R
function/procedure), 109, 134, 154,
156 radio, 174, 317, 323, 362
procedure definition, 108–109, 154, 158 radix sorting, 283
processing unit, 273 raster graphics (see graphics, raster),
189
professional ethics, 295, 331–335
reasoning by cases, 63–68
program verification (see verification,
reasoning by induction, 247–253
program), 161, 263
recipe, 15, 34, 235
programmable digital computer, 142 recursion, 281
programmer, 11–12, 15, 22, 36–37, 45, recursive problem solving, 281
48–51, 59, 61–62, 85, 98, 109, 125, reliable code, 117
131, 147–148, 151, 169, 201, repetition, 25–30, 62, 249, 281
207–208, 219, 224–225, 259, 272, return, 144, 146–147, 150–151, 169
283, 335, 342–343, 355, 363 rigorous argument (see argument,
programming error (see bug), 37, 41, rigorous), 24, 53–68, 91, 153–161,
48–50, 61–62, 98, 125, 161 255, 355
programming language, 11, 15, 28, robot, 34, 79–84, 148, 235, 345–352,
31–37, 62, 114, 125, 148, 161, 235, 358
358 robotics, 34, 148, 235, 345–352, 358,
programming language, choosing, 28, 361
36, 148 robotics, biologically inspired, 351
programming language, functional, robotics, soft, 345, 351
148 Roman numerals, 171–174, 179
programming language, high-level, 37
programming language, low-level, 37 S
programming language, object-oriented, safety-critical programs, 141, 255
235 safety-critical software, 141, 255
programming language, universal, 36 sand board, 12
378 Conjuring with Computation

scale, 120, 359 subroutine (see function, procedure,

Scratch, 11 method), 147
search, 30, 109, 147, 169, 272, 298, surveillance, 320, 322
335, 355, 363 surveillance society, 319–323
search algorithm (see algorithm, swap (code), 101–104, 127–130
search), 169, 355 swap (data), 64–68, 80–82, 111–118,
secure operating system (see operating 203–206, 212–213
system, secure), 314 swarm computing, 351
security, 114, 295–335, 360 symbol, 1, 125, 174, 184, 186, 300,
security level, 311, 315 302, 308, 356
security protocol, 114 system image, 224–225
selection, 19–25, 28, 62–63, 65–67,
138–139, 184, 194, 333 T
semaphore, 307 tally, 173
sequence, functions, 149–151 telegraph, 174
sequence of instructions, 13–17, 22–25, test, in a program, 19, 22, 25, 28–29
28, 149, 151, 153–161, 225, 234, 355 testing, 41–51, 53, 56–57, 59–62, 259,
sequence, procedures, 153–161 263, 355
sequencing, 13–17, 24–25, 28, 153–161 testing, end-to-end, 51
shared editing, 235 topology, 15
shopping website, 214, 235, 242, 359 trojan horse program, 322
true value (see boolean value), 22–24,
signifier, 357
28
smart glasses, 342–343
trusted person, 315
social context, 357
Turing complete, 28
social engineering, 328
Turing machine, 4, 148
social network (see network, social), Turing test, 4
288, 329
social psychology, 357 U
software engineering, 51, 242, 359
unary, 173
software, safety critical, 141, 255
undo, 207, 232
sort, 109, 147, 169, 282–283
unfulfilled needs, 243
sort algorithm (see algorithm, sort),
universal programming language (see
169, 283
programming language, universal), 36
spaghetti code, 62 usability, 199, 201–209, 219, 221–227,
specification, 104, 112–114, 129, 134, 243
137, 140–141, 154, 158, 161 usability design principle (see design
specification language, 104, 141 principle), 79, 84, 215, 217, 219,
statement, 22–24, 62, 65, 67 226, 357
steganography, 297–302, 307–308, 311, usability engineering, 219
315–316, 360 usability evaluation methods (see
Strasbourg Airport, 218 evaluation methods, usability), 219
string, datatype, 134 usability guideline (see design
string, knotted (see Quipu), 168 principle), 79, 84, 215, 217, 219,
structured programming, 28, 62 226, 357
 Index of Computing Terms 379

usability principle (see design principle), vector graphics (see graphics, vector),
79, 84, 215, 217, 219, 226, 357 189
usable, 73, 84–85, 199, 208, 219, 237, verification, program, 161, 263
357 virtual world, 339–343
used, 207, 357 visibility, 215–219
user-centred design, 219 visibility of system state/status,
user experience, 199, 237–244, 357, 215–219
359 vision system, 71–74, 211–214
user experience design, 199, 237–244, visual design language, 94
357, 359 visual perception, 71–74, 211–214
user interface, 73, 77, 93–94, 207, 235 visual salience, 211–214
user interface designer, 207 voice authentication (see
user interface, graphical (GUI), 73, 77, authentication, voice), 363
93–94, 235 von Neumann architecture, 273
user model (see model, user),
224–225 W
UX design (see user experience design), wearable computing, 360
199, 237–244, 357, 359 white hat hacker (see hacker), 331, 335
wireless communications, 115, 323
V wireless control, 115, 323
valve, 142 wizard, programming, 12
variable, 124, 259, 263 word processor, 208
This page intentionally left blank
 Index of Tricks, Illusions and
Conjuring Techniques

(*) means a trick that is described in the book in detail.

A cognitive psychology, 69–85, 199–244,

Afghan Bands (*), 17 356–358
Communicating secretly by code,
alchemists, 34
165–170, 181–197, 297–323
Are You Psychic (*), 237–244, 247–253
Create a Forcing Matrix (*) (see
Forcing Matrix), 143–148, 149–150
B
crooked casino games, 110
Barnum statements (*), 325–330 Cyclic False Cut (*), 90, 131–134, 141,
238–241, 250, 252
C Cyclops’ Eye (*), 71–74
Calculator Conjuring (*), 31–37
Call the Clairvoyant (*), 311–317 D
Call the Clairvoyant a Second Time Domino trick (*), 288–289, 293–294
(*), 319–323 Doomsday Clock (*), 267–273, 289
Cards appearing and disappearing from Double Destination Deception (*),
fingertips, 120 285–294
card sharp, 130, 227 Down-Under Deal and Robotic Reveal
card trick, 7–12, 25–30, 53–68, (*), 345–352
101–134, 120, 143–148, 153–161, Drawn to You (*), 165–170
190–197, 201–214, 217, 221–264,
275–283, 297–302, 303–309, E
311–317, 345–352, 366 Egg of Columbus, 115
Cards on your Mind (*), 181–186 Ends Overhand Shuffle (*), 117–119
Catching bullets, 253 exposing frauds, 130, 330, 334–335
cheating at card games, 110, 130, 227
Chevreul Pendulum (*), 331–335 F
Chopping off your head, 309 Faerie Cage (*), 339–343
Classic Mentalism (*), 303–309 False Choice (*), 19–24, 143–152,
close-up magic, 227, 238, 244 204–205

381
382 Conjuring with Computation

False Cut (*) (see Cyclic False Cut, Misdirection, 203–206, 209, 222,
Three Way False Cut), 90, 131–134, 232–233
141, 221–227, 238–241, 250, 252 Möbius Rings (*), 13–17
False Shuffle (*) (see False Cut), 105, Most Expensive Trick in the World, 30
111–130, 132, 141, 146, 250
False Top Overhand Shuffle (*), N
121–125, 129, 146, 230, 232, 347, Naming the dead (*), 170
349 Noah’s ark, 253
Find the Lady, 63 Numbers Game (*), 215–219
Floating lightbulb, 17
Forced Wizardly Book Magic (*), O
149–152
Forcing Matrix (*), 143–148 On the Power of Numbers (*), 187–190
Four Aces (*), 201–209 Op Art, 74
Free My Three (*), 275–283, 347 One True Sovereign (*), 353–363
optical illusion, 71–74, 170
G Out-of-Body Experience (*),
191–197
ghostly figures or images (*), 330, 332,
Overhand Shuffle (*), 107–110, 255,
339–343
258, 278, 297–298, 300
Overhand Shuffle (variations),
I
111–130, 146, 230, 232, 347, 349
Invisible Palming (*), 7–12, 101–103,
153–161 P
Paranormal, 335
J
Passing a bottle through a table,
Joker in the Pack (*), 229–235 244
performance (see showmanship), 1,
L 96–100, 152, 204, 230, 294, 334,
Levitation, 130 351–352, 355, 359, 363
Lottery Trick (*), 175–179 Pledge, The, 99–100
low gravity magic, 363 Pointillism, 74, 94
Pre-booked Picture Magic (*), 89–94
M Prestige, The, 99–100
magical objects, 13–17, 19–24, 41–45, Prestige, The (film), 115
71–77, 79–85, 95–104, 115, 149–154, Prestige, The (novel), 99
237–244, 267–273, 319–323, Psychometry, 170
331–335, 339–343, 348, 353–363 Pulling rabbits out of a top hat, 152
Magically Weighted Boxes (*), 75–77
Make coins appear, 152 R
Mechanical Turk, 348 Red-Black Mind Meld (*), 255–264
Mental Monte and the Trains of Riffle shuffle, 110, 358
Thought (*), 63–68 Rising card, 120
Mentalism, 170, 174, 303–330, 360 Roman Maths Challenge (*),
Mind reading, 53–62, 89–94, 181–186 171–174
 Index of Tricks, Illusions and Conjuring Techniques 383

S T
Seance, 100, 130, 331, 334, 343 21-Card Trick (*), 53–62
self-working trick, 10–12, 15, 45, 83, Telepathy, 317
244, 258, 354, 358 Teleporting Robot (*), 79–85,
showmanship, 351, 355, 359, 363 361
shuffling (*), 107–130, 146, 230, 232, Teleporting Top Hat (*), 101–104,
255, 258, 278, 297–298, 300, 347, 153–154
349 Three Card Monte, 63
Side Jogging (*), 127–130 Three Way False Cut (*),
Six Magical Objects (*), 19–24 221–227
Sniff Out that Card (*), 297–302 Time misdirection, 204, 222,
spiritualism, 100, 130, 191–197, 328, 232–233
330–335, 343 Trained Rice (*), 95–100
street magic, 30, 152, 244, 294 Trick gone wrong ending, 26, 30
supernatural powers, 100, 130, Turn, The, 99–100
211–214, 237–264, 334–335 Turning the Key (*), 25–30
Supernatural Suggestion (*), 211–214
Superpower strength, 77 W
Swapped Ends Overhand Shuffle (*), Wizardly Book Magic (*), 41–51, 56,
111–115, 117–118 149–152, 270, 272
This page intentionally left blank
 Index of People

A Eisenhower, General, 142

Aldrin, Buzz, 51 Ellis, Clarence, 235
Antheil, George, 323 Elmsley, Alex, 283
Aristotle, 24 Euler, Leonhard, 291–293
Arthur, King, 353
F
B Faraday, Michael, 100, 294
Babbage, Charles, 125, 180, 186 Flowers, Tommy, 142
Bacon, Sir Francis, 300, 302 Floyd, Robert W., 161
Barnum, Phineas T., 326–328, 330 Frayne, Steven
Bayer, David, 110 (see Dynamo), 294
Bell, Tim, 197
Blackstone Sr., Harry, 17 G
Boole, George, 24
Gardner, Martin, 68
Bowie, David, 115
Garriott, Richard, 348, 363
Buchanan, Joseph, 170
Goldstine, Herman, 273
C Goldwasser, Shafi, 264
Gorillaz, 342
Church, Lorenzo, 148
Clarke, Arthur C., 1
H
D Hamilton, Margaret, 51
Darwin, Charles, 130 Herrmann, Adelaide, 253
Devant, David, 152 Hoare, Tony, 161
Diaconis, Persi, 110 Hooper, William, 85
Dijkstra, Edsger W., 62 Hopper, Grace, 37, 49
Dircks, Henry, 343 Hummer, Bob, 68
Dynamo, 294 Hurst, Lulu, 77
Hutchins, Edwin, 356
E
I
Eckert, J. Presper, 273
Edison, Thomas, 17 Ingalls, Dan, 235

385
386 Conjuring with Computation

J Papert, Seymour, 104

Jastrow, Joseph, 170 Pendragon, Uther, 353
al-Jazari, Ismail, 352 Penn and Teller (see Jillette, Penn;
Jillette, Penn, 30 Teller, Raymond), 30
Pepper, John Henry, 343
K Piddington, Lesley, 317
Piddington, Sydney, 317
Kare, Susan, 94
della Porta, Giambattista, 343
Kay, Alan, 235
Presto, Fay, 244
von Kempelen, Wolfgang, 348
al-Khwārizmī, Muhammad ibn Mūsā,
R
12
Kordemsky, Boris A., 85 Rackoff, Charles, 264
Kuhn, Gustav, 209 Randi, James, 335
Riley, Bridget, 74
L Robert-Houdin, Jean-Eugène, 3,
348–349, 351, 362
Lamarr, Hedy, 323
Leibniz, Gottfried, 186
S
Lloyd, Sam, 85
Lovelace, Ada, 125 Scot, Reginald, 309
Seurat, Georges, 74, 94
M Shakespeare, William, 42, 309
Smith, David, 235
Madonna, 342
Snowden, Edward, 315–316
Marconi, Guglielmo, 174
Soyel, Hamit, 214
Mary, Queen of Scots, 300
Maskelyne, John Nevil, 130, 174
T
Maskelyne, Nevil, 174
Mauchly, John, 273 Teller, Raymond, 30
McCarthy, John, 148 Tesla, Nikola, 115
McOwan, Peter W., 214, 348 Tesler, Larry, 235
McQueen, Steve, 227 Thurston, Howard, 120
Merlin, 353 Trewey, Felicien, 17
Micali, Silvio, 264 Turing, Alan, 4, 28, 142, 148

N V
von Neumann, John, 273 Verma, Milan, 214
Nielsen, Jakob, 219 da Vinci, Leonardo, 190
Nolan, Christopher, 115
Norman, Don, 219 W
Wallace, Alfred Russel, 130
O Wheeler, David, 134
Ose, Jay, 227 Wilkes, Maurice, 134
Ôuchi, Hajime, 72, 74 Wing, Jeanette, 104

P X
Pacioli, Luca, 190 Xerox PARC Team, 235