LOGICAL INFERENCE

&
PROOFs
Debdeep Mukhopadhyay
Dept of CSE, IIT Madras

Defn
A theorem is a mathematical assertion
which can be shown to be true. A proof is
an argument which establishes the truth of
a theorem.

Nature & Importance of Proofs

In mathematics, a proof is:
a correct (well-reasoned, logically valid) and complete
(clear, detailed) argument that rigorously &
undeniably establishes the truth of a mathematical
statement.

Why must the argument be correct & complete?

Correctness prevents us from fooling ourselves.
Completeness allows anyone to verify the result.

In this course (& throughout mathematics), a

very high standard for correctness and
completeness of proofs is demanded!!

Overview
Methods of mathematical argument (i.e.,
proof methods) can be formalized in terms
of rules of logical inference.
Mathematical proofs can themselves be
represented formally as discrete
structures.
We will review both correct & fallacious
inference rules, & several proof methods.

Applications of Proofs
An exercise in clear communication of logical
arguments in any area of study.
The fundamental activity of mathematics is
the discovery and elucidation, through
proofs, of interesting new theorems.
Theorem-proving has applications in
program verification, computer security,
automated reasoning systems, etc.
Proving a theorem allows us to rely upon on
its correctness even in the most critical
scenarios.

Proof Terminology
Theorem
A statement that has been proven to be
true.

Axioms, postulates, hypotheses,

premises
Assumptions (often unproven) defining the
structures about which we are reasoning.

Rules of inference
Patterns of logically valid deductions from
hypotheses to conclusions.

More Proof Terminology

Lemma - A minor theorem used as a steppingstone to proving a major theorem.
Corollary - A minor theorem proved as an easy
consequence of a major theorem.
Conjecture - A statement whose truth value has
not been proven. (A conjecture may be widely
believed to be true, regardless.)
Theory The set of all theorems that can be
proven from a given set of axioms.

Graphical Visualization
A Particular Theory

A proof

The Axioms
of the Theory

Various Theorems

Inference Rules - General Form

An Inference Rule is
A pattern establishing that if we know that a
set of antecedent statements of certain forms
are all true, then we can validly deduce that a
certain related consequent statement is true.

antecedent 1
antecedent 2
consequent
therefore

means

Inference Rules & Implications

Each valid logical inference rule
corresponds to an implication that is a
tautology.
antecedent 1
Inference rule
antecedent 2
consequent
Corresponding tautology:
((ante. 1) (ante. 2) ) consequent

Some Inference Rules

p
pq
pq
p

p
q
pq

Rule of Addition
Rule of Simplification
Rule of Conjunction

Modus Ponens & Tollens

p
pq
q
q
pq
p

Rule of modus ponens

(a.k.a. law of detachment)
the mode of
affirming

Rule of modus tollens

the mode of denying

Syllogism Inference Rules

pq
qr
pr
pq
p
q

Rule of hypothetical
syllogism
Rule of disjunctive
syllogism

Aristotle
(ca. 384-322 B.C.)

Formal Proofs
A formal proof of a conclusion C, given
premises p1, p2,,pn consists of a
sequence of steps, each of which applies
some inference rule to premises or
previously-proven statements
(antecedents) to yield a new true
statement (the consequent).
A proof demonstrates that if the premises
are true, then the conclusion is true.

Formal Proof Example

Suppose we have the following premises:
It is not sunny and it is cold.
We will swim only if it is sunny.
If we do not swim, then we will canoe.
If we canoe, then we will be home early.
Given these premises, prove the theorem
We will be home early using inference rules.

Proof Example cont.

Let us adopt the following abbreviations:
sunny = It is sunny; cold = It is cold;
swim = We will swim; canoe = We will
canoe; early = We will be home early.

Then, the premises can be written as:

(1) sunny cold (2) swim sunny
(3) swim canoe (4) canoe early

Proof Example cont.

Step
1. sunny cold
2. sunny
3. swimsunny
4. swim
5. swimcanoe
6. canoe
7. canoeearly
8. early

Proved by
Premise #1.
Simplification of 1.
Premise #2.
Modus tollens on 2,3.
Premise #3.
Modus ponens on 4,5.
Premise #4.
Modus ponens on 6,7.

Inference Rules for Quantifiers

x P(x)
P(o)
P(g)
x P(x)
x P(x)
P(c)
P(o)
x P(x)

(substitute any specific object o)

(for g a general element of u.d.)

(substitute a new constant c)

(substitute any extant object o)

Common Fallacies
A fallacy is an inference rule or other
proof method that is not logically valid.
A fallacy may yield a false conclusion!

Fallacy of affirming the conclusion:

pq is true, and q is true, so p must be
true. (No, because FT is true.)
If he stole, he will be nervous when he is
interrogated. He was nervous when
interrogated, so he stole.

Fallacy
Fallacy of denying the hypothesis:
pq is true, and p is false, so q must be
false. (No, again because FT is true.)
If his hands are full of blood, he has
murdered. But he is sitting on his sofa, well
dressed (without any sign of blood), so he did
not murder.
He may have washed his hands !!!

Slightly complicated example

Statement:

x[P(x) Q(x)] xP(x) xQ(x)

Quick Check: P(x): x is even, Q(x): x is odd

Fallacious Proof:
x [P(x) Q(x)] x [P(x) Q(x)]
Remember we
x[P(x) Q(x)]
Proved in the last
[x P(x) x Q(x)]
class
[ x P(x) x Q(x)]
xP(x) xQ(x)
Fallacy of denying the antecedent

Circular Reasoning
The fallacy of (explicitly or implicitly)
assuming the very statement you are trying to
prove in the course of its proof. Example:
Prove that an integer n is even, if n2 is even.
Attempted proof: Assume n2 is even.
Then n2=2k for some integer k. Dividing both
sides by n gives n = (2k)/n = 2(k/n). So there
is an integer j (namely k/n) such that n=2j.
Therefore n is even.
How do
show in
that
j=k/n=n/2
is an integer,
Circular reasoningyou
is used
this
proof. Where?
without first assuming that n is even?

A Correct Proof
We know that n must be either odd or even.
If n were odd, then n2 would be odd, since
an odd number times an odd number is
always an odd number. Since n2 is even, it
is not odd, since no even number is also an
odd number. Thus, by modus tollens, n is
not odd either. Thus, by disjunctive
syllogism, n must be even.
This proof is correct, but not quite complete,
since we used several lemmas without proving
them. Can you identify what they are?

A More Verbose Version

Suppose n2 is even 2|n2 n2 mod 2 = 0.
Of course n mod 2 is either 0 or 1.
If its 1, then n1 (mod 2), so n21 (mod 2)
Now n21 (mod 2) implies that n2 mod 2 = 1.
So by the hypothetical syllogism rule,
(n mod 2 = 1) implies (n2 mod 2 = 1).

Since we know n2 mod 2 = 0 1, by modus

tollens we know that n mod 2 1.
So by disjunctive syllogism we have that
n mod 2 = 0 2|n n is even. Q.E.D.

Proof Methods for Implications

For proving implications pq, we have:
Direct proof: Assume p is true, and prove
q.
Indirect proof: Assume q, and prove p.
Vacuous proof: Prove p by itself.
Trivial proof: Prove q by itself.
Proof by cases:
Show p(a b), and (aq) and (bq).

Direct Proof Example

Definition: An integer n is called odd iff n=2k+1
for some integer k; n is even iff n=2k for some k.
Theorem: (For all numbers n) If n is an odd
integer, then n2 is an odd integer.
Proof: If n is odd, then n = 2k+1 for some
integer k. Thus, n2 = (2k+1)2 = 4k2 + 4k + 1 =
2(2k2 + 2k) + 1. Therefore n2 is of the form 2j +
1 (with j the integer 2k2 + 2k), thus n2 is odd.

Indirect Proof Example

Theorem: (For all integers n)
If 3n+2 is odd, then n is odd.
Proof: Suppose that the conclusion is false, i.e.,
that n is even. Then n=2k for some integer k.
Then 3n+2 = 3(2k)+2 = 6k+2 = 2(3k+1). Thus
3n+2 is even, because it equals 2j for integer j =
3k+1. So 3n+2 is not odd. We have shown that
(n is odd)(3n+2 is odd), thus its contrapositive (3n+2 is odd) (n is odd) is also true.

Vacuous Proof Example

Theorem: (For all n) If n is both odd and
even, then n2 = n + n.
Proof: The statement n is both odd and
even is necessarily false, since no
number can be both odd and even. So,
the theorem is vacuously true.

Trivial Proof Example

Theorem: (For integers n) If n is the sum
of two prime numbers, then either n is odd
or n is even.
Proof: Any integer n is either odd or
even. So the conclusion of the implication
is true regardless of the truth of the
antecedent. Thus the implication is true
trivially.

Proof by Contradiction
A method for proving p.
Assume p, and prove both q and q for
some proposition q. (Can be anything!)
Thus p (q q)
(q q) is a trivial contradiction, equal to
F
Thus pF, which is only true if p=F
Thus p is true.

Proof by Contradiction Example

Theorem:

2 is irrational.

Proof: Assume 21/2 were rational. This

means there are integers i,j with no common
divisors such that 21/2 = i/j. Squaring both
sides, 2 = i2/j2, so 2j2 = i2. So i2 is even; thus i
is even. Let i=2k. So 2j2 = (2k)2 = 4k2.
Dividing both sides by 2, j2 = 2k2. Thus j2 is
even, so j is even. But then i and j have a
common divisor, namely 2, so we have a
contradiction.

Review: Proof Methods So Far

Direct, indirect, vacuous, and trivial proofs
of statements of the form pq.
Proof by contradiction of any statements.
Next: Constructive and nonconstructive
existence proofs.

Proving Existentials
A proof of a statement of the form x P(x)
is called an existence proof.
If the proof demonstrates how to actually
find or construct a specific element a such
that P(a) is true, then it is a constructive
proof.
Otherwise, it is nonconstructive.

Constructive Existence Proof

Theorem: There exists a positive integer n
that is the sum of two perfect cubes in two
different ways:
equal to j3 + k3 and l3 + m3 where j, k, l, m are
positive integers, and {j,k} {l,m}

Proof: Consider n = 1729, j = 9, k = 10,

l = 1, m = 12. Now just check that the
equalities hold.

Another Constructive
Existence Proof
Theorem: For any integer n>0, there
exists a sequence of n consecutive
composite integers.
Same statement in predicate logic:
n>0 x i (1in)(x+i is composite)
Proof follows on next slide

The proof...

Given n>0, let x = (n + 1)! + 1.

Let i 1 and i n, and consider x+i.
Note x+i = (n + 1)! + (i + 1).
Note (i+1)|(n+1)!, since 2 i+1 n+1.
Also (i+1)|(i+1). So, (i+1)|(x+i).
x+i is composite.
n x 1in : x+i is composite. Q.E.D.

Nonconstructive Existence
Proof
Principle of extremum
Theorem:
There are infinitely many prime numbers.
Any finite set of numbers must contain a
maximal element, so we can prove the
theorem if we can just show that there is no
largest prime number.
i.e., show that for any prime number, there is
a larger number that is also prime.
More generally: For any number, a larger
prime.
Formally: Show n p>n : p is prime.

The proof, using proof by

cases...
Given n>0, prove there is a prime p>n.
Consider x = n!+1. Since x>1, we know
(x is prime)(x is composite).
Case 1: x is prime. Obviously x>n, so let
p=x and were done.
Case 2: x has a prime factor p. But if pn,
then x mod p = 1. So p>n, and were
done.

Proof by contradiction
Assume a largest prime number exists; call it p.
Form the product of the finite number of prime
numbers,
r=2.3.5.7p
Now inspect r+1: It cannot be divisible by any of
the above prime numbers
So, either r+1 is a prime or divisible by a prime
greater than p (There is a fallacy in Stanats proof).
Thus, in either case there is a prime greater than
p, and hence we have a contradiction
Thus, there is no maximum prime number and
the set is infinite.

Adaptive proofs
Adapt the previous proof to prove that
there are infinite prime numbers of the
form 4k+3, where k is a non-negative
integer.

The Halting Problem (Turing36)

The halting problem was the first
mathematical function proven to
have no algorithm that computes it!
We say, it is uncomputable.

The desired function is Halts(P,I) :

the truth value of this statement:
Program P, given input I, eventually terminates.

Theorem: Halts is uncomputable!

I.e., There does not exist any algorithm A that
computes Halts correctly for all possible inputs.

Its proof is thus a non-existence proof.

Corollary: General impossibility of predictive
analysis of arbitrary computer programs.

Alan Turing
1912-1954

The Proof

Absurd makes a
liar out of HALT, by
doing the opposite
of whatever HALT
predicts.

Given any arbitrary program HALT(P)

Consider algorithm Absurd, defined as:
procedure Absurd:
if HALT(Absurd)==T
while T begin end

Note that Absurd halts iff

H(Absurd) = F.
So H does not compute the function
Halts!

Limits on Proofs
Some very simple statements of number
theory havent been proved or disproved!
E.g. Goldbachs conjecture: Every integer n2
is exactly the average of some two primes.
n2 primes p,q: n=(p+q)/2.

There are true statements of number

theory (or any sufficiently powerful system)
that can never be proved (or disproved)
(Gdel).