Thinking 2 - Heuristics! The Representativeness Heuristic!

Exemplar 1! Exemplar 2!

Exemplar 3! Exemplar 4!

• Representativeness heuristic is related to the idea of exemplar-

based categorization.!
• If you’re asked to determine whether this image is a cat, you
Danny Kahneman! Amos Tversky!
https://www.princeton.edu/~kahneman/! http://news.stanford.edu/pr/96/960605tversky.html! can compare it with your exemplars of the cat category to see
if this new image seems representative of this category.!

The Representativeness Heuristic! The Representativeness Heuristic!

• Most people answer that H T T H H T H T looks

A coin is tossed 8 times -- more random and is therefore more likely to
Which of the following is occur by chance.!
more likely to occur?!
• However, H T T H H T H T and H H H H H H H H
HHHHHHHH or! are equally likely.!
H T T H H T H T! – There are 256 different sequences of 8 coin flips, and
each of these sequences is equally likely to occur.!

– H T T H H T H T has a 1/256 chance of occurring.!

– So does H H H H H H H H.!

The Representativeness Heuristic! The Representativeness Heuristic!

• People typically view this as Exemplar 1! Exemplar 2!

a simple categorization
problem.!
• They compare these two
sequences to other Exemplar 3! Exemplar 4!

instances of random
sequences that they’ve
seen.!
• H T T H H T H T looks more
like a representative • Our brains are built to do this kind of categorization
instance of the category of rapidly, and it works well for things like cats.!
random sequences! • It takes a lot more work to go through the math to figure
• This is fast and easy, but out probabilities for things like flipping a coin.!
leads to an error in this task!

© S. J. Luck

All rights reserved
1

 The Representativeness Heuristic! Analogical Mapping!

• One of the keys to solving a

problem is seeing analogies
between the problem and other
problems that you’ve previously
solved.!

• Also, if you’re not trained in probability theory, you might • Imagine that a child is given the task of figuring out how
think that this looks like a simple categorization problem.! to divide this 24-ounce pitcher of orange juice equally
• One of the keys to solving a problem like this is seeing among four people.!
analogies between the problem and other problems that • How many ounces of juice does each person get?!
you’ve previously solved.!

Analogical Mapping! Analogical Mapping!

• Seems easy to you because

you see that it is analogous
to other division problems.!
• 24 ÷ 4 = 6!
• But not easy for a young
child who has little 24 ÷ 4 = 6!

experience with division! • You might have previously encountered a situation in which you had to
distribution a pile of M&Ms among 4 people.!
• You could do this by counting them out – one for you, one for you, one for
you, etc.!
• And each person would end up with 6 M&Ms!
• The same approach could be taken with the 24 ounces of juice.!
24 ÷ 4 = 6! • But this would require seeing that the juice problem and the M&M problem are
analogous, which requires seeing the mapping between an ounce of juice and
a single M&M.!

The Representativeness Heuristic!

• Research by cognitive psychologists has shown

that people who solve a problem are usually the
people who figured out the correct analogy with
another problem.!
• When people come up with the wrong analogy,
they come up with a wrong solution.! • The gambler’s fallacy: Believing that the next event is
more likely to be a winner if you’ve had several
• People fail to answer the coin toss question
consecutive losses!
correctly because they think of this as a
categorization problem, which leads them to use • The fact that you’ve already lost 3 games can’t really
representativeness.! influence what happens on the next game, at least not in
a typical poker game.!
• The answer instead requires thinking about
each possible sequence of coin tosses and the • The cards that are left might differ in the 4th game, but
probability of each sequence.! that influences everybody equally and doesn’t change
your chances of winning or losing.!

© S. J. Luck

All rights reserved
2

 Win! Lose! Lose! Lose! Lose!
Lose! Win! Lose! Lose! Lose!
Lose! Lose! Win! Lose! Lose!
Lose! Lose! Lose! Win! Lose!

Win! Lose! Lose! Lose! Lose!

Lose! Win! Lose! Lose! Lose!
Lose! Lose! Win! Lose! Lose!
Lose! Lose! Lose! Win! Lose!
• When looking at this problem, people think about the fact
that losing 4 out of 4 games is more probable than losing • Losing 3 games in a row doesn’t increase the probability
that you’ll win in the 4th game.!
3 out of 4 games.!
• This is true, because there are 4 ways to lose 3 out of 4
games, but only one way to lose 4 out of 4.!
• However, the outcome of the 4th game is independent of
what happened in the previous 3 games.!

The Representativeness Heuristic! The Representativeness Heuristic!

Bill is 34 years old. He is intelligent, but
unimaginative, compulsive, and generally
lifeless. In school, he was strong in math
but weak in social studies and
humanities.!
Which of the following is most probable?!
A) Bill plays jazz for a hobby!
B) Bill is an accountant who plays jazz
for a hobby!
Subjects were asked: !
They choose B because Bill doesn’t seem very
Which of the following is most probable?! representative of the category of people who play jazz for a
A) Bill plays jazz for a hobby! hobby.!

B) Bill is an accountant who plays jazz for a hobby! But he does seem representative of the category of
accountants.!
>80% answer that B is more likely than A! Kahneman & Tversky (1982)! Kahneman & Tversky (1982)!

The Representativeness Heuristic! All people!

Category A-!
People who play
jazz for a hobby!
• It can’t be true that it’s more likely that Bill is an
accountant who plays jazz for a hobby than it is Category C-!
Accountants!
that he plays jazz for a hobby.!
• All accountants who play jazz for a hobby are Category B-!
Accountants who play
people who play jazz for a hobby.!
jazz for a hobby!
• A subset can’t be more likely than the larger set All accountants who play jazz for a hobby are members of
the group of people who play jazz for a hobby; therefore,
that contains it.! it cannot be more likely that someone is an accountant
who plays jazz for a hobby than that someone is a person
who plays jazz for a hobby!

• This shows, out of all people, the set of people who play
jazz for a hobby, the set of people who are accountants,
and the intersection of these two sets, which is the set of
accountants who play jazz for a hobby.!

© S. J. Luck

All rights reserved
3

 All people! The Availability Heuristic!
Category A-!
People who play
jazz for a hobby!
• The availability heuristic: when you’re asked to judge
Category C-! how common something is, you do this by searching
Accountants!
your memory for instances of it.!
Category B-! • If you can think of a lot of instances, you think that
Accountants who play
thing must be quite common.!
jazz for a hobby!
All accountants who play jazz for a hobby are members of • That’s a very reasonable heuristic, but it assumes that
the group of people who play jazz for a hobby; therefore,
it cannot be more likely that someone is an accountant the number of times you hear about something is a
who plays jazz for a hobby than that someone is a person good measure of how often it occurs, which is not
who plays jazz for a hobby!
always the case.!

p(A) = # of people in A ÷ total # of people in population!

p(B) = # of people in B ÷ total # of people in population!
Because B < A, p(B) < p(A)!

Diabetes versus Homicide! Breast Cancer versus Prostate Cancer!

Breast Cancer: 72,815 cases! Prostate Cancer: 218,890 cases!

I think ______ cancer occurs more commonly than
______ cancer by a factor of ______.!

Diabetes = 72,815! Homicide = 16,692! • People hear a lot more about breast cancer than about
(2005)! (2005)!
prostate cancer, so they don’t realize that prostate
• People hear about homicide deaths more than diabetes
cancer is 3 times as common as breast cancer.!
deaths, so they don’t realize that diabetes deaths are
more common !

The Availability Heuristic! The Anchoring-and-Adjustment

Heuristic
Which is more common:!
!
• When you’re asked to guess a quantitative
A) Words that begin with the letter R! value, like how many countries are in Africa, you
B) Words that have R as the third letter! need some kind of reference point.!
• When I asked you how many countries are in
Radio! Perfect! Africa, I mentioned that it’s easy to think of 5:!
Relative! Bird!
– Kenya, Libya, South Africa, Egypt, Ethiopia!
Rug! Car!
Ridiculous! ???! • That biased you to use 5 as the anchor.!
Rascal! • The problem is that people tend not to adjust
Reputation! far enough from the starting point.!

© S. J. Luck

All rights reserved
4

 Africa! The Anchoring-and-Adjustment Heuristic!

• If I had instead pointed out that there are ~195 countries

in the world, you probably would have overestimated the
# of countries in Africa.!
• You would have used 195 as your anchor, and adjusted
I think there are ______ countries in Africa.!
downward from there.!

The Anchoring-and-Adjustment Heuristic! Base Rate Neglect!

• When I asked you about
the population of
England, I started by • A base rate is how often
noting that the something would be
population of the U.S. is a expected, independent of
little over 300 million.! specific information that
we’re given.!
• That gave you an anchor
point, and you adjusted
• If a doctor is trying to diagnose a disease on the basis of
downward from 300
an X ray, the base rate would be the general prevalence
million.!
of the disease in the population.!
• Most people don’t adjust
far enough, so this leads • If a disease is very rare, then a given person probably
them to overestimate.! doesn’t have it even if the X ray provides some indication
of this rare disease.!
I think the population of England is ______.!

Bayes’s Theorem!
P(H|E) = P(H) x P(E|H)! • We need to take into account the
! !! P(E)! prior probability that the person
has the disease (the probability of
the disease before we see the
evidence from the X ray).!
• Best estimate of the prior
probability is the overall
Bayes’s Theorem! prevalence of the disease in the
population.!
P(H|E) = P(H) x P(E|H)!
Thomas Bayes (1701–1761)! • If the disease is rare (prior
! !! P(E)!
probability is low), then this
• In the X ray example, we’re is trying to figure out the decreases the probability that the
likelihood that someone has a disease given the evidence Thomas Bayes (1701–1761)! person has the disease, even if
from an X ray.! the X ray is consistent with that
disease.!

© S. J. Luck

All rights reserved
5

 Base Rate Neglect! Base Rate Neglect!
Base rate information:!
209 Engineering professors!
I have a friend who is a professor. She
likes poetry, traveling, and fine wines. 9 Art History professors!
Is she in the Art History Department • This doesn’t take into account the fact that there are 20 times as
or in the College of Engineering?! many engineering professors as art history professors at UC Davis.!
• Things like poetry, traveling, and wine might be slightly more likely
Most people think she’s likely to be in
to be true of a given art history professor than a given
Art History, because things like engineering professor,!
poetry, travel, and wine are more
• The fact that there are 20 times as many engineering professors
representative of our concept of an
as art history professors is going to have a much bigger impact on
art history professor than our the likelihood that someone is an engineering vs. art history
concept of an engineering professor! professor.!
• People tend to ignore these base rates, especially when they’re
not made obvious.!
Kahneman & Tversky (1973)!

Examples: Peet’s Disease and Starbukine! • Background information!

– 1 in 1000 women who gets a
mammogram has Peet’s
Disease!
– Mammograms detect 99% of
Peet’s Disease cases!
– Only 0.5% of mammograms
are false positives!
• Patient Z.B. has a
mammogram and tests
positive for Peet’s Disease!
– What is the probability that
she actually has it?!
– Wrong answer: 99.5%!

• Wrong answer: her probability of having Peet’s disease is 99.5%!

– Based on the idea that only .5% of positive tests are false positives, so
there is only a .5% chance of that someone with a positive result is
• These are completely made-up examples so that we can having a false positive.!
have exact numbers.! • But that doesn’t take into account the very low base rate that
someone actually has Peet’s disease.!

• Background information! • Background information!

– 1 in 1000 women who gets a – 1 in 1000 women who gets a
mammogram has Peet’s mammogram has Peet’s
Disease! Disease!
– Mammograms detect 99% of – Mammograms detect 99% of
Peet’s Disease cases! Peet’s Disease cases!
– Only 0.5% of mammograms – Only 0.5% of mammograms
are false positives! are false positives!
• Patient Z.B. has a • Patient Z.B. has a
mammogram and tests mammogram and tests
In 100,000 positive for Peet’s Disease! In 100,000 positive for Peet’s Disease!
mammograms! – What is the probability that mammograms! – What is the probability that
she actually has it?! she actually has it?!
Has Peet’s! No Peet’s! Has Peet’s! No Peet’s!
– Wrong answer: 99.5%! – Wrong answer: 99.5%!
Positive 99! 500! Positive 99! 500! – Only 99 of 600 women with a
Test Result! Test Result! positive test actually have
Peet’s Disease!
Negative 1! 99400! Negative 1! 99400!
Test Result! Test Result! – Z.B.’s chance of having Peet’s
Disease is only ~1 out of 6!

© S. J. Luck

All rights reserved
6

 • Judgment!
– Drawing conclusions about facts!
– Which brands of cars are most reliable?!
• Decision-making!
– Deciding what to do!
– Which car should I buy?!
• Background Information!
– Starbukine cures Peet’s disease in 80% of women who
have the disease and take Starbukine! • You might get the
– 1% of women who take Starbukine experience a life- impression that people
threatening side effect! are lousy at making
• Should all women over age 30 take Starbukine?! judgments and
• If base rate of Peet’s Disease is 1/1000, and decisions.!
100,000 women take Starbukine! • But that’s not the point
– Of the 100 who actually had the disease, 80 will be cured! of this research.!
– 1,000 otherwise healthy women will experience the life-
threatening side effects of Starbukine!

Descriptive Models! Bounded Rationality!

• Humans have limited time and mental abilities, so they’re

bound to make some mistakes.! • People around the world make decisions about when to
cross a street millions of times a day.!
• But in many real-world cases people usually make the
right decisions.! • They almost always make the right decision.!

Herb Simon!
http://diva.library.cmu.edu/webapp/simon/!

Analogical Mapping!

24 ÷ 4 = 6!

• When people make mistakes, often it’s because they

haven’t figured out the right analogy to other
problems.!
• Sometimes decisions reflect a competition between a
• They rely on things like representativeness and slow, rational mechanism and a fast, emotional
availability in situations where they ought to think mechanism.!
more carefully about the nature of the problem.!

© S. J. Luck

All rights reserved
7

 The Representativeness Heuristic! Dual Process Theories!

• Dual Process Theories of Judgment & Decision-Making.!

• Your mind can take two different approaches to solving a
• This isn’t just an issue of rationality versus emotion.! problem:!
• It’s an issue of making a quick decision versus thinking
• A slow, careful, controlled approach!
slowly and carefully.!
• A fast and more automatic approach!
• For example, if you tell people to think carefully about all
the different ways that a coin can be flipped 8 times, • The fast approach isn’t always more emotion-based than
they’re more likely to come up with the right answer to the slow approach, but often it is.!
this question.!

The Instance Theory of Automaticity! Dual Process Theories!

• Instance theory of automaticity:!

• Two pathways for performing a task (e.g.,
arithmetic):! • Similarly, dual-process theories of JDM assume that we
– Automatic memory retrieval of previous instances:
Presentation of the stimulus leads to a memory search (e.g., have controlled and automatic processes that compete
“16 + 16 = 32”)! for control over behavior.!
– Controlled process: Use a rule to perform the task (“Add 6 +
6, carry the one, …”)!
• These pathways race, and whichever finishes first
controls behavioral output!

Key Issues in Cognitive Psychology

✔!Architecture of the human mind!

Limits on human abilities (e.g., speed, capacity)!
Representations (format, persistence)!
Processing Steps / Algorithms!
Hardware (cognitive neuroscience)!
Differences among healthy individuals, across • Two competing views of Language:!
development, and in disorders! • The mind contains special-purpose modules for things like
Real-world applications! language acquisition!
• Language acquisition is achieved by domain-general
mechanisms.!
• Is the mind divided into special-purpose modules for things like
language, face recognition, spatial processing, etc.?!

© S. J. Luck

All rights reserved
8

 Syntax! Semantics!
Key Issues in Cognitive Psychology
The principles that govern Meanings of utterances
the arrangement of (morphemes, phrases,
linguistic structures within sentences, paragraphs, etc.)!
and between sentences! Architecture of the human mind!
Limits on human abilities (e.g., speed, capacity)!
✔!Representations (format, persistence)!
Processing Steps / Algorithms!
Hardware (cognitive neuroscience)!
Differences among healthy individuals, across
development, and in disorders!
Real-world applications!

"Syntax must be bad, having both

sin and tax in it.” -Will Rogers!

Analog vs. Propositional

Representations! PO
VIP

MST 7a
Analog representations:! Propositional representations:! MT
Quantitative physical Arbitrary relationship LIP
variations in the between the form of the Retina Thalamus V1
representation map onto representation and the thing V3
quantitative physical V2
being represented!
variations in the thing ! V4 TEO
being represented! !
! ! TF TE
Example: Representing the Example: Representing the
quantity eighteen as:! quantity eighteen as:!
“18” or “XVIII”!

Analog! Propositional!

Networks of Concepts!
Key Issues in Cognitive Psychology

Architecture of the human mind!

Limits on human abilities (e.g., speed, capacity)!
Representations (format, persistence)!
✔!Processing Steps / Algorithms!
Hardware (cognitive neuroscience)!
Differences among healthy individuals, across
development, and in disorders!
Real-world applications!

© S. J. Luck

All rights reserved
9

 Concepts & Categories:
A Synthesis! Key Issues in Cognitive Psychology

Architecture of the human mind!

Limits on human abilities (e.g., speed, capacity)!
Instances! Classical!
View! Representations (format, persistence)!
Processing Steps / Algorithms!
✔!Hardware (cognitive neuroscience)!
Differences among healthy individuals, across
development, and in disorders!
Real-world applications!

Prototypes! Theories!

Mental Imagery and Area V1! Key Issues in Cognitive Psychology

Perception! Imagery!
Architecture of the human mind!
Horizontal vs. Vertical!

Limits on human abilities (e.g., speed, capacity)!

Representations (format, persistence)!
Processing Steps / Algorithms!
Hardware (cognitive neuroscience)!
Differences among healthy individuals, across
development, and in disorders!
Area V1! Area V1!
✔!Real-world applications!

(Klein et al., 2004)!

• Research on judgment & decision-making has had a big

impact on understanding how medical and economic
decisions are actually made, including the errors and
biases that often occur when people make judgments and
decisions.!

• This has led to changes in the ways that doctors make

diagnoses and select treatments!

© S. J. Luck

All rights reserved
10