Artificial Intelligence

Automated Reasoning
Andrea Torsello

Automated Reasoning
Very important area of AI research

Reasoning usually means deductive reasoning

New facts are deduced logically from old ones

Inductive reasoning (later in course)

Guessing facts from old ones and from evidence

Two main aspects of deductive reasoning

Logical representations (thousands of them)

Rules of deduction (how to deduce new things)

Applications of Automated Reasoning

Automated mathematics

Axioms(A) are given, theorem statement(T) is given

Reasoning agent searches from A to T (or from T to A)

Using rules of deduction to move around the search space

Automated verification

Hardware and Software verification

That they perform as specified

Remember the Intel chip fiasco?

Intel now have lots of people working on automated verification

Automating Deductive Reasoning

Deduce new knowledge from old

Prove/disprove some open conjectures

Theorem proving

Search for a path from axioms to theorem statement

Operators are (sound) inference rules

Applications:

Agents that use deductive inference

Mechanising and automating mathematics

Verifying hardware and software specifications

The semantic web

Inference Rules
A entails B iff

B is true when A is true

Any model of A is a model of B

Then this is a sound inference rule

A
B
Axioms C D Z Theorem

Each step is application of inference rule

Theorem is entailed by the axioms

Tautologies
S: (X(YZ))((XY) (XZ))

Show that no matter what truth values for X, Y and Z

The statement S is always true

YZ

XY

XZ

X (YZ)

((XY)(XZ))

true

true

true

true

true

true

true

true

true

true

true

false

false

true

false

false

false

true

true

false

true

false

false

true

false

false

true

true

false

false

false

false

false

false

false

true

false

true

true

true

true

true

true

true

true

false

true

false

false

true

true

true

true

true

false

false

true

false

true

true

true

true

true

false

false

false

false

true

true

true

true

true

Columns 7 and 8 are always the same

Inference with Tautologies

Regardless of meaning or truth values of P and Q

This is content-free and a tautology

One way to define a rule of inference:

We can replace PQ with QP, and vice versa

They are true for same models

Replacing one for other preserves soundness

Equivalence Rules
A and B are logically equivalent (write A B)

Same models for each

Can replace any instance of A with an instance of B without affecting models

Formalised as rewrite rule A B

Also B A

Must avoid looping A B A B ...

Choose one direction, or always loop-check

Rewrite rules used for inference

Showing theorem and axioms are equivalent

Preprocessing theorem/axioms into a particular format

Properties
Commutativity

PQ QP
PQ QP
PQ QP
Associativity

(PQ)R P(QR)
(PQ)R P(QR)
Distributivity over 'and' and 'or':

P (Q R) (P Q) (P R)
P (Q R) (P Q) (P R)
Distributivity over implication

P (Q R) (P Q) (P R)
P (Q R) (P Q) (P R)

Properties
De Morgans Law (refers to either)

(P Q)

P Q

(P Q)

P Q

Contraposition: imagine the opposite is true

PQ

Q P

Often useful in mathematics proof

Remove implication or equivalence (very useful)

P Q P Q
P Q (P Q) (Q P)
Reduce to truth value

P P False
P P True

An Example Deduction
(P Q) (P Q)
Show that this sentence is false

Show that this rewrites to False

This proves the negation

1. Using the double negation rewrite: P => P

(P Q) (P Q)

2. Using De Morgan's Law: P Q => (P Q)

(P Q) (P Q)

3. Using the commutativity of : P Q => Q P

(P Q) (Q P)

4. Using 'replace implication' from right to left: P Q => P Q

(P Q) (Q P)

5. Using 'replace equivalence' from left to right: P Q => (P Q) (Q P)

((P Q) (Q P)) (Q P)

6. Using the associativity of : (P Q) R => P (Q R)

(P Q) ((Q P) (Q P))

7. Using the consistency equivalence above: P P => False

(P Q) False

8. Using the definition of :

False

Propositional Inference Rules

Rewrite rules are good for bidirectional search

But we dont need equivalence, just entailment

Classic example

All men are mortal, socrates is a man

Therefore: Socrates is mortal

This is an instance of an inference rule

Known as Modus Ponens (Aristotle)

AB, A
B
Above line: what we know, below: what we can deduce

Soundness of Modus Ponens

AB

Top: AB, A

Bottom: B

True

True

True

True

True

True

False

False

False

False

False

True

True

False

True

False

False

True

False

True

And-Elimination & -Introduction

And-Elimination:
A1 A 2 A n
Ai
[1 i n]

And-Introduction:
A1, A2, , An
A1 A2 An
The sentences may be from different places

Selected from the database

Or-Introduction & Unit Resolution

Ai
A1 A2 An
[1 i n]

Unit resolution

(A B) , B
A
Basis for resolution theorem proving

Substitution & Instantiation

Substitute into a variable by assigning a particular value

Replace with given term, remove quantifier

Instantiation (grounding) is a kind of substitution

Must substitute a ground term

Example: X.Y.likes(X,Y)

becomes likes(tony, george)
We write:

Subst({X/tony, Y/george}, likes(X,Y)) = likes(tony,george)

Universal Elimination
Given a sentence, A

Containing a universally quantified variable V

Then we can replace V by any ground term g

V.A
Subst({V/g}, A)
Remember to remove quantifier

Not as complicated as it looks:

X likes(X, ice_cream)

becomes likes(ben,ice_cream)

Existential Elimination
Given a sentence, A

Containing an existentially quantified variable, V

Then we can replace V by any constant, k

As long as k is not mentioned anywhere else

V.A
Subst({V/k}, A)
For the sake of argument, lets call it

Existential Introduction
Given a sentence, A

And a variable, V, which is not used in A

Then any ground term, g, in A can be substituted by V

As long as g does not appear in A also

A
V. Subst({g/V},A)
Exercise: find sentence where V is in A such that this inference rule is not sound

Chains of Inference

Remember the problem were trying to solve

Search for a path from axioms, A, to theorem, T

Three approaches

Forward chaining

Backward chaining

Proof by contradiction

Specification of a search problem:

Representation of states (first-order logic sentences)

Initial state (depends...)

Operators (inference rules, including equivalences)

Goal state (depends...)

Forward Chaining

Deduce new facts from axioms

Deduce new facts from these, etc.,

Hopefully end up deducing the theorem statement

A1

A2

A3

Backward Chaining
Start with the theorem state and work backwards

Hope to end up at the axioms

Each step asks: given the state that Im at...

Which operator could have been applied to which state to produce the state
(sentence) Im at

No problem when using equivalences

Can also use a bidirectional search (from both ends)

Difficult when using general inference rules

Many possible ways to invert operators

Reductio ad absurdum
Assume theorem is false

then show that the assumption contradicts the axioms

which proves that the theorem is true

Add negated theorem to the set of axioms

See if we can deduce the False sentence

Advantage: using the theorem statement from start

Can look to see how close we are to the false statement

Possibilities for a heuristic search!

Basis for resolution theorem proving

(next week)