Default Reasoning

By
Naval Chopra(07010015)
Pranay Bhatia (07005005)
Pradeep Kumar(07D05020)
Siddharth Chinoy(07D05005)
Vaibhav Chhimpa(07D05011)
Motivation

The problem of what to do next

Implicit assumptions, rules of thumb,
inferential shortcuts

Default reasoning : "Inference to the
first unchallenged alternative"

Is this reasoning? Yes, it often leads to
useful knowledge.

Real people, unlike philosophers, have
to make judgments and decisions all the
time, not just when they're good and
ready
 Look before you leap - A proverb
He who hesitates is lost - Another proverb


Reasoning

Thinking in the pattern of an argument

Moving from premise to conclusion

People are generally capable of good reasoning

When people make mistakes, they are due to
deviations from the ideal

The 'closer' we are to following the logically valid
pattern of argument, the better the reasoning

Wrong!

Courtesy : Kent Bach
Introduction
Default reasoning is a form of
Defeasible Reasoning used to
express facts like “by default,
something is true”.
Default Logic is a Non-Monotonic
Logic proposed by Raymond
Reiter to formalize reasoning with
default assumptions.
Introduction
Standard logic can only express
that something is true or that
something is false.
This is a problem because
reasoning often involves facts
that are true in the majority of
cases but not always.
Example : “Birds typically fly” vs
“All birds fly”
Exceptions – Penguins, Ostriches
Defeasible Reasoning
Reasoning is defeasible when the
corresponding argument is
rationally compelling but not
deductively valid.
The truth of the premises of a
good defeasible argument
provide support for the
conclusion, even though it is
possible for the premises to be
true and the conclusion false.
Defeasible Reasoning
Inother words, the support provided
by the premises to the conclusion is a
tentative one, potentially defeated by
additional information.
Defeasible reasoning has typically
been limited to inferences involving
exception-permitting generalizations,
that is, inferring what has or will
happen on the basis of what normally
happens.
Problems with First Order
Logic
First Order Logic is a Monotonic
Logic in the sense that its
consequence relation is
monotonic.
If a sentence φ can be inferred in
FOL from a set Γ of premises,
then it can also be inferred from
any set Δ of premises containing
Γ as a subset.
Problems with First Order
Logic
i.eThe consequence relation of
FOL has the property that if Γ ⊨ φ
and Γ ⊆ Δ then Δ ⊨ φ. This
property is known as the
Monotonic Property.
Intuitively, this implies that
learning a new piece of
knowledge cannot reduce the set
of what is known.
Problems with First Order
Logic
There are striking differences between
formal logic and the working of the
mind when it comes to dealing with
Incomplete Knowledge (such as
perception, ambiguity, common sense,
causality and prediction)
Classical logic lacks tools for
describing how to revise a formal
theory to deal with inconsistencies
caused by new information.
Problems with First Order
Logic
There are mainly two types of
revisions :
World Model Reorganisation – Very
hard problem of revising complex
models. Complexity usually stems
from part of the model relying on
other parts of the model.
Eg. Revision of one’s opinion of a
friend after discovering his dishonesty
Problems with First Order
Logic
Routine Revision – Easier
Problem. Involves maintaining
facts which although expressed
as universally true have
exceptions.
Eg. Stating that all birds usually
fly, but then on finding out that
penguins don’t fly revising the
knowledge base to include that
fact.
Problems with First Order
Logic
Classical logic has overlooked the
above simple cases by altering
the notation in which rules are
stated.
Eg. “All birds except Penguins,
Ostriches, ... Fly”
Classical logic also cannot handle
abductive reasoning
(consequences deduced as most
likely explanations).
Default Reasoning
Default Reasoning (and Default
Logic) was proposed to handle
the problems of non-monotonicity
and belief revision.
It mainly aims at formalising
default inference rules without
stating all the exceptions.
Logics formalizing default
reasoning
Logics able to deal with arbitrary
default assumptions (default
logic, defeasible reasoning, and
answer set programming)
Logics that formalize the specific
default assumption that facts
that are not known to be true can
be assumed false by default
(closed world assumption and
circumscription).
Syntax of Default Logic
A default theory is a pair <D,W>
W is a set of logical formulae,
called the background theory,
that formalize the facts that are
known for sure.
D is a set of default rules, each
one being of the form:
Prerequisite : Justification1,
…,Justificationn
Conclusion
Syntax of Default Logic
According to this default, if we
believe that Prerequisite is true,
and each of Justificationi is
consistent with our current
beliefs, we are led to believe that
Conclusion is true.
The logical formulae in W and all
formulae in a default were
originally assumed to be first-
order logic formulae, but they
can potentially be formulae in an
arbitrary formal logic.
Examples for Syntax
The default for “Birds typically fly” is
formalised by the following default :

The rule means that if X is a bird and it

can be assumed that it flies, then we
conclude that it flies

A background theory for the above is :

Example for Syntax

What we can conclude :

Flies(Condor)

What we cannot conclude :

Flies(Penguin)

Bird(Eagle)
 Courtesy : http://www.rci.rutgers.edu/~cfs/472_html/Logic_KR/DefaultTheory.html

This default rule is applicable if we

can prove from our beliefs that John
is an American and an adult, and
believing that there is some car
that is owned by John does not lead
to an inconsistency.
If these two sets of premises are
satisfied, then the rule states that
we can conclude that John owns a
car.
Another Example
A common default assumption is
that what is not known to be true
is believed to be false.
This is known as the Closed
World Assumption, and is
formalized in default logic using a
default like the following one for
every fact F.
Restrictions
A default is categorical or
prerequisite-free if it has no
prerequisite (or it’s prerequisite is
tautological).
A default is normal if it has a
single justification that is
equivalent to its conclusion.
A default is supernormal if it is
both categorical and normal.
A default is seminormal if all its
justifications entail its conclusion.
Semantics of Default Logic
A default rule can be applied to a
theory if its precondition is entailed
by the theory and its justifications
are all consistent with the theory.

When the theory is such that no

other default can be applied, the
theory is called an extension of the
default theory.
The default rules may be applied in
different order, and this may lead to
different extensions.
Nixon Diamond
Usually, Quakers are pacifist
Usually, Republicans are not
pacifist
Richard Nixon is both a Quaker
and a Republican

Courtesy : http://www.cs.cf.ac.uk/Dave/AI2/node80.html
Nixon Diamond
Since Nixon is a Quaker, one
could assume that he is a
pacifist; since he is Republican,
however, one could also assume
he is not a pacifist.
Entailment
Entailment of a formula from a
default theory can be defined in two
ways:
◦ Skeptical
 A formula is entailed iff it is entailed by all its
extensions
 Since Nixon can neither be proved to be a pacifist
nor the contrary, no conclusion is drawn.
◦ Credulous
 A formula is entailed iff it is entailed by at
least one of its extensions.
 Since Nixon can be proved to be a pacifist in at
least one case, he is believed to be a pacifist;
however, since he can also be proved not be a
pacifist, he is also believed not to be a pacifist.
Some examples
The following default theory has
no extension:

Normal default theory (At least

one extension present):
Alternate Default Inference
Rules
Justified - differs from the
original one in that a default is
not applied if thereby the set T
becomes inconsistent with a
justification of an applied default
Concise - a default is applied
only if its consequence is not
already entailed by T (the exact
definition is more complicated
than this one; this is only the
main idea behind it)
Alternate Inference Rules
Constrained - a default is applied
only if the set composed of the
background theory, the justifications
of all applied defaults, and the
consequences of all applied defaults
(including this one) is consistent.
Rational - similar to constrained
default logic, but the consequence of
the default to add is not considered in
the consistency check
Alternate Default Inference
Rules
Cautious - defaults that can be
applied but are conflicting with
each other (like the ones of the
Nixon diamond example) are not
applied.
Variations
 Priorities on defaults
◦ Relative priority of defaults are explicitly specified. Among the defaults that
are applicable to a theory, only one of the most preferred one is applied.
◦ One form of implicit priority : more specific defaults (those that are
applicable in fewer cases) are preferred over less specific ones.

http://www.rci.rutgers.edu/~cfs/472_html/Logic_KR/KnowledgeRepToc.html
Variations
Statistical variant
◦ Statistical default is a default with an
attached upper bound on its
frequency of error; in other words, the
default is assumed to be an incorrect
inference rule in at most that fraction
of times it is applied.
◦ Extensions for statistical default logic
are constructed in the usual way,
except that the operator ‘terminates’
when inference reaches a specified
threshold.
Variations
 Weak extensions
◦ Rather than checking whether the preconditions are valid in the
theory composed of the background theory and the
consequences of the applied defaults, the preconditions are
checked for validity in the extension that will be generated.
◦ In other words, the algorithm for generating extensions starts by
guessing a theory and using it in place of the background theory;
what results from the process of extension generation is actually
an extension only if it is equivalent to the theory guessed at the
beginning.
◦ This variant of default logic is related in principle
to autoepistemic logic, where a theory has a model in which x is
true just because, assuming true, the formula supports the initial
assumption.
◦ A logic allowing such a self-support of beliefs is called not
strongly grounded to differentiate them from strongly
grounded logics, in which self-support is not possible. Strongly
grounded variants of autoepistemic logic exist. <Get more info>
Circumscription
Non-monotonic logic that formalizes
the common sense assumption that
things are as expected unless
otherwise specified.
In the Missionaries and Cannibals
Problem, the solution “go half a
mile south and cross the river on
the bridge” is intuitively not valid
The problem statement does not
mention a bridge.
Does not exclude it’s existence
either.
Circumscription in FOL
Given a FOL formula T containing a
predicate P, circumscribing this
predicate amounts to selecting only
the models of T in which P is
assigned to true on a minimal set of
tuples.
The extension of a predicate in a
model is a set of tuples that this
predicate assigns to true in that
model
The circumscription of a predicate P
in a formula T is obtained by
selecting only the models of T with
a minimal extension of P.
Circumscription in FOL

Where p<P =
 p is a predicate having the same arity as P
and x is an n-tuple
 The above formula means that there exists
no predicate p which assigns false to every
value that P assigns false and is not P
 By adding an extra literal ~Abnormal(..) to
each fact stating that it holds only in
normal conditions.
 Minimizing the extension allows for
reasoning under the implicit assumption
that things are as expected and that this
assumption is made only if possible
Closed World Assumption
The presumption that what is not
currently known to be true, is false.
Negation as failure is related as it
amounts to believing false every
predicate that cannot be proved to be
true.
In formal logic it is achieved by adding
to the knowledge base the negation of
the literals that are not currently
entailed by it
Resultant system may not be
consistent
 Implementations
DeReS, XRay, GADel

XRay : An experimental theorem

prover for query-answering from
incomplete knowledge bases.
Implementation relies heavily on
Mark Stickel's Prolog Technology
Theorem Prover, PTTP.

S. Brüning and T. Schaub. A model-based approach to consistency-checking.

Implementations

◦ Deres - written in C, run under Unix

◦ Relaxed stratification - Divide and
conquer approach to computing
extensions
◦ Original default theory partitioned
into several sub theories called strata
◦ Extensions of theory are recomputed
via the extensions of the strata.
◦ Efficient, when strata are small.
  The intermediate language - PTTP for formulas in negation
normal form. Here's its grammar:

 An example file :
% facts
adult.
% default rules
adult :- student : adult.
not_employed :- student : not_employed.
not_married :- student : not_married.
employed:-adult:not_student.
married:-adult:not_student.
% initial query
query:-married;employed.

M. Stickel. A Prolog technology theorem prover. New Generation Computing, 2:371-383, 1984.
 References
1. M. Stickel. A Prolog technology theorem prover. New Generation
Computing, 2:371-383, 1984.
2. S. Brüning and T. Schaub. A model-based approach to consistency-
checking. In Z. Ras and M. Michalewicz, editors, Proceedings of the
Ninth International Symposium on Methodologies for Intelligent
Systems, volume 1079 of Lecture Notes in Artificial Intelligence,
pages 315-324. Springer Verlag, 1996.
3. Elaine Rich & Kevin Knight, "Artificial Intelligence", McGraw-Hill
Science/Engineering/Math; 2nd edition.
4. Russel S. and Norvig P., "Artificial Intelligence: a Modern
Approach", Prentice Hall, 1998.
5. Nilsson, N.J., "Artificial Intelligence, a New Approach", Morgan
Kaufmann, 2000.
6. http://www.cs.cf.ac.uk/Dave/AI2/node74.html
7. http://plato.stanford.edu/entries/reasoning-defeasible/
8. http://plato.stanford.edu/entries/logic-nonmonotonic/
9. http://online.sfsu.edu/~kbach/defaultreasoning.pdf
10.http://www.cs.cmu.edu/~jgc/publication/
Default_Reasoning_Inheritance_Mechanisms_SIGART_1980.pdf
Thank you!
Questions ?