Introduction: Mathematical

Paradoxes

Intuitive approach. Until recently, till the end

of the 19th century, mathematical theories
used to be built in an intuitive or axiomatic
way.

The historical development of mathematics

has shown that it is not sufficient to base
theories on an intuitive understanding
of their notions only.

This fact became especially obvious in set

theory.

1
The basic concept of a set is certainly taken
from reality, for there we come across many
examples of various sets, all of which are
finite.

But in mathematics it is also necessary to

consider infinite sets, such as the set of
all integers, the set of all rational numbers,
the set of all segments, the set of all tri-
angles.

2
Intuitively, by a set, we mean any collection
of objects- for example, the set of all even
integers or the set of all students in a class.

The objects that make up a set are called its

members (elements).

Sets may themselves be members of sets for

example, the set of all sets of integers has
sets as its members.

3
Most sets are not members of themselves.

For example the set of all students is not a

member of itself, because the set of all
students is not a student.

However, there may be sets that do belong

to themselves.

For example the set of all sets.

A following simple reasoning indicates that it

is necessary to impose some limitations
on the concept of a set.

4
Russell Paradox, 1902: Consider the
set A of all those sets X such that X is
not a member of X.

Clearly, A is a member of A if and only if

A is not a member of A.

So, if A is a member of A, the A is also

not a member of A; and

if A is not a member of A, then A is a

member of A.

In any case, A is a member of A and A is

not a member of A.

CONTRADICTION

5
Russell solution: every object must have a
definite non-negative integer as its type.
An expression x is a member of the set
y is meaningful if and only if the type of
y is one greater than the type of x.

Theory of types says that it is meaningless

to say that a set belongs to itself, there
can not be such set A, as stated in the
Russell paradox.

Development: by Whitehead and Russell in

years 1910 - 1913. It is successful, but
difficult in practice and has certain other
drawbacks as well.

6
 LOGICAL PARADOXES or
ANTINOMIES

Logical Paradoxes (antinomies) are the para-

doxes concerning the notion of a set.

Axiomatic Set Theory is a general solution

to Logical Paradoxes.

It is one of the most important fields of Mod-

ern Mathematics or more specifically Math-
ematical Logic and Foundations of Math-
ematics.

Zermello, 1908 - first paradoxes free axiomatic

set theory.

7
Two of the other most known logical para-
doxes are Cantor and Burali-Forti antinomies.
They were stated at the end of 19th cen-
tury.

Cantor paradox involves the theory of cardi-

nal numbers.

Burali-Forti paradox is the analogue to Can-

tor’s in the theory of ordinal numbers.

8
Cardinal number: cardX = cardY or X and
Y are equinumerous if and only if there
is one-to-one correspondence that maps X
and Y ).

cardX ≤ cardY means that

X is equinumerous with a subset of Y .

cardX < cardY means that

cardX ≤ cardY and cardX 6= cardY .

9
Cantor Theorem: For any set X,

cardX < cardP(X).

Schröder- Berstein Theorem: For any sets

X and Y ,

If cardX ≤ cardY and cardY ≤ cardX, then

cardX = cardY .

Ordinal numbers are the numbers assigned

to sets in a similar way as cardinal num-
bers but they deal with ordered sets.

10
Cantor Paradox, 1899: Let C be the uni-
versal set - that is, the set of all sets.
Now, P(C) is a subset of C, so it follows
easily that

cardP(C) ≤ cardC.

On the other hand, by Cantor theorem,

cardC < cardP(C) ≤ cardP(C), so also

cardC ≤ cardP(C)
and by Schröder- Berstein theorem we have
that
cardP(C) = cardC,
what contradicts Cantor Theorem: cardC <
cardP(C).

Solution: Universal set does not exist.

11
Burali-Forti Paradox, 1897 Given any ordi-
nal number, there is a still larger ordinal
number. But the ordinal number deter-
mined by the set of all ordinal numbers is
the largest ordinal number.

Solution: set of all ordinal numbers do not ex-

ist.

12
Semantic Paradoxes another solution: reject
the assumption that for every property P (x),
there exists a corresponding set of all ob-
jects x that satisfy P (x).

Russell’s Paradox then simply proves that there

is no set A of all sets that do not belong
to themselves.

Cantor Paradox shows that there is no uni-

versal set.

Burali-Forti Paradox shows that there is no

set that contains all ordinal numbers.

13
A more radical interpretation of the paradoxes
has been advocated by Brower and his in-
tuitionist school.

Intuitionists refuse to accept the universality

of certain basic logical laws, such as the
law of excluded middle: A or not A.

For intuitionists excluded middle law is true

for finite sets, but it is invalid to extend it
to all sets.

The intuitionists’ concept of infinite set dif-

fers from that of classical mathematicians.

14
Infinite set for the intuitionists is something
which is constantly in a state of formation.

Example: set of positive integers is infinite be-

cause to any given finite set of positive in-
tegers it is always possible to add one more
positive integer.

For intuitionists The notion of the set of all

subsets of the set of all positive integers is
not regarded meaningful.

Intuitionists’ mathematics is different from

that of most mathematicians in their re-
search.

The basic difference lies in the interpreta-

tion of the word exists.

15
Example: let P (n) be a statement in the
arithmetic of positive integers. For the
mathematician the sentence the exists n,
such that P(n) is true if it can be de-
duced (proved) from the axioms of arith-
metic by means of classical logic.

In classical mathematics proving existence

of x does not mean that one is able to
indicate a method of construction of a
positive integer n such that P (n) holds.

For the intuitionist the sentence the exists

n, such that P(n) is true only if he is able
to construct a number n such that P (n)
is true.

16
In general in the intuitionists’ universe we are
justified in asserting the existence of an ob-
ject having a certain property only if we
know an effective method for constructing
or finding such an object.

In intuitionist’ mathematics the paradoxes are,

in this case, not derivable (or even mean-
ingful).

Intuitionism because of its constructive fla-

vor, has found a lot of applications in com-
puter science, for example theory of pro-
grams correctness.

Intuitionists logic reflects intuitionists ideas in

a form formalized deductive system.

17
The axiomatic theories solved some but not
all problems.

Consistent set of axioms does not prevent the

occurrence of another kind of paradoxes,
called Semantic Paradoxes.

18
 SEMANTIC PARADOXES

Berry Paradox, 1906: Let

A denote the set of all positive inte-

gers which can be defined in the En-
glish language by means of a sentence
containing at most 1000 letters.

The set A is finite since the set of all sen-

tences containing at most 1000 letters is
finite.
Hence, there exist positive integer which
do not belong to A.

19
The sentence:

n is the least positive integer which

cannot be defined by means of a sen-
tence of the English language contain-
ing at most 1000 letters

contains less than 1000 letters and defines

a positive integer n.

Therefore n belongs to A. But n does not

belong to A by the definition of n.

CONTRADICTION!

20
Berry Paradox Analysis: The paradox resulted
entirely from the fact that we did not say
precisely what notions and sentences be-
long to the arithmetic and what notions
and sentences concern the arithmetic,
examined as a fix and closed deductive sys-
tem.

And on the top of it we also mixed the nat-

ural language with mathematical lan-
guage.

21
Solution: (Tarski) we must always
distinguish the language of the theory
we talk about (arithmetic) and the lan-
guage which talks about the theory,
called a metalanguage.

In general we must distinguish a theory from

the meta-theory.

22
The Liar Paradox (Greek philosopher Eubu-
lides of Miletus, 400 BC)

A man says: I am lying.

If he is lying, then what he says is true, and

so he is not lying.

If he is not lying, then what he says is not

true, and so he is lying.

In any case, he is lying and he is not lying.

CONTRADICTION

23
Löb Paradox (1955 )

Let A be any sentence. Let B be a sentence:

If this sentence is true, then A. So, B
asserts: If B is true then A.

Now consider the following argument: As-

sume B is true. Then, by B, since B is
true, A is true.

This argument shows that, if B is true, then

A. But this is exactly what B asserts. Hence,
B is true. Therefore, by B, since B is true,
A is true. Thus every sentence is true.

CONTRADICTION

24
TARSKI solution: these paradoxes arise be-
cause the concepts of ” I am true”, this
sentence is true, ” I am lying” should not
occur in the language (theory). It belong
to a metalanguage (meta-theory).

25
The Liar Paradox is a corrected version of a
following paradox stated in antiquity by a
Cretan philosopher Epimenides.

Cretan ” Paradox” (The Cretan philosopher

Epimenides paradox, 600 BC)

Epimenides, a Cretan said: All Cretans

are liars. If what he said is true, then,
since Epimenides is a Cretan, it must be
false. Hence, what he said is false. Thus,
there must be some Cretan who is not
a liar.

Note that the conclusion that there must be

some Cretan who is not a liar is not logi-
cally impossible, so we do not have a gen-
uine paradox.

26
 General Remarks

First task of building mathematical logic, foun-

dations of mathematics or computer sci-
ence is to define their symbolic language.
This is called a syntax.

Second task is to extend the syntax to in-

clude a notion of a proof. It allows us
to find out what can and cannot be proved
if certain axioms and rules of inference are
assumed. This part of syntax is called a
proof theory.

27
Third task is to define what does it mean
that formulas of our language are true,
i.e. to define a semantics for the lan-
guage.

For example the notion of truth for classi-

cal and intuitionistic approaches are differ-
ent; classical and intuitionistic semantics
are different.

Fourth task is to investigate the relationship

between proof theory (part of the syntax)
and semantics.

This relationship is being established by prov-

ing fundamental theorems:
Soundness, Completeness, Consistency.

28
Role of Classical and Non-classical Logics
in Computer Science

Classical Logic The use of classical logic on

computer science is known, undisputable,
and well established. The existence of PRO-
LOG and Logic Programming as a separate
field of computer science is the best exam-
ple of it.

Intuitionistic Logic in the form of Martin-Löf’s

theory of types (1982), provides a com-
plete theory of the process of program spec-
ification, construction , and verification. A
similar theme has been developed by Con-
stable (1971) and Beeson (1983).

29
Modal Logic: In 1918, an American philoso-
pher, C.I. Lewis proposed yet another in-
terpretation of lasting consequences of the
logical implication. In an attempt to avoid,
what some felt, the paradoxes of implica-
tion (a false sentence implies any sen-
tence) he created a modal logic.

The idea was to distinguish two sorts of truth:

necessary truth and a mere possible (con-
tingent) truth.

A possibly true sentence is one which, though

true, could be false.

30
Modal Logic in Computer Science is used as
as a tool for analyzing such notions as knowl-
edge, belief, tense.

Modal logic has been also employed in form

of Dynamic logic (Harel 1979) to facilitate
the statement and proof of properties of
programs.

31
Temporal Logics were created for the speci-
fication and verification of concurrent pro-
grams (Harel, Parikh, 1979, 1983), for a
specification of hardware circuits Halpern,
Manna and Maszkowski, (1983), to specify
and clarify the concept of causation and its
role in commonsense reasoning (Shoham,
1988).

Fuzzy logic, Many valued logics were created

and developed to reasoning with incom-
plete information.

32
The development of different logics and the
applications of logics to different areas of
computer science or even artificial intelli-
gence only is beyond the scope of class.

In class we will define and study the basic prop-

erties of some of the most standard non-
classical logics: many valued, intuition-
istic, and modal.

33
 Computer Science Puzzles
Reasoning in Distributed Systems

Grey, 1978, Halpern, Moses, 1984: Two di-

visions of an army are camped on two hill-
tops overlooking a common valley. In the
valley awaits the enemy.

If both divisions attack the enemy simulta-

neously they will win the battle.

If only one division attacks it will be defeated.

The divisions do not initially have plans for

launching an attack on the enemy, and the
commanding general of the first division
wishes to coordinate a simultaneous attack
(at some time the next day).

34
Neither general will decide to attack unless
he is sure that the other will attack with
him.

The generals can only communicate by means

of a messenger.

Normally, it takes a messenger one hour to

get from one encampment to the other.

However, it is possible that he will get lost

in the dark or, worst yet, be captured by
the enemy.

Fortunately, on this particular night, every-

thing goes smoothly.

Question: How long will it take them to co-

ordinate an attack?

35
Suppose the messenger sent by General A makes
it to General B with a message saying At-
tack at dawn.

Will B attack? No, since A does not know B

got the message, and thus may not attack.

General B sends the messenger back with an

acknowledgment. Suppose the messenger
makes it.

Will A attack? No, because now A is worried

that B does not know A got the message,
so that B thinks A may think that B did
not get the original message, and thus not
attack.

36
General A sends the messenger back with an
acknowledgment.

This is not enough. No amount of acknowl-

edgments sent back and forth will ever guar-
antee agreement. Even in a case that the
messenger succeeds in delivering the mes-
sage every time. All that is required in this
(informal) reasoning is the possibility that
the messenger doesn’t succeed.

37
Solution: Halpern and Moses (1985) created
a Propositional Modal logic with m
agents. They proved this logic to be es-
sentially a multi-agent version of the modal
logic S5.

They proved that common knowledge (for-

mally defined!) not attainable in systems
where communication is not guaranteed

Also it is also not attainable in systems where

communication is guaranteed, as long as
there is some uncertainty in massage de-
livery time.

38
In distributed systems where communication
is not guaranteed common knowledge is
not attainable.

But we often do reach agreement!

They proved that common knowledge (as for-

mally defined) is attainable in such models
of reality where we assume, for example,
events can be guaranteed to happen simul-
taneously.

Moreover , there are some variants of the

definition of common knowledge that are
attainable under more reasonable assump-
tions.

So, we can prove that in fact we often do

reach agreement!

39
 Reasoning in Artificial Intelligence

Flexibility of reasoning is one of the key prop-

erty of intelligence.

Commonsense inference is defeasible in its

nature; we are all capable of drawing con-
clusions, acting on them, and then retract-
ing them if necessary in the face of new
evidence.

If Computer programs are to act intelligently,

they will need to be similarly flexible.

Goal: development of formal systems that de-

scribe commonsense flexibility.

40
 Flexible Reasoning Examples

Reiter, 1987 Consider a statement Birds fly.

Tweety, we are told, is a bird. From this,
and the fact that birds fly, we conclude that
Tweety can fly.

This is defeasible: Tweety may be an ostrich,

a penguin, a bird with a broken wing, or a
bird whose feet have been set in concrete.

Non-monotonic Inference: on learning a new

fact (that Tweety has a broken wing), we
are forced to retract our conclusion (that
he could fly).

41
Non-monotonic Logic is a logic in which the
introduction of a new information (ax-
ioms) can invalidate old theorems.

Default reasoning (logics) means drawing of

plausible inferences from less-then- con-
clusive evidence in the absence of infor-
mation to the contrary.

Non-monotonic reasoning is an example of

the default reasoning

42
Moore, 1983 Consider my reason for believ-
ing that I do not have an older brother. It is
surely not that one of my parents once ca-
sually remarked, You know, you don’t have
any older brothers, nor have I pieced it to-
gether by carefully sifting other evidence.

I simply believe that if I did have an older

brother I would know about it;

therefore, since I don’t know of any older broth-

ers of mine, I must not have any.

43
”The brother” reasoning is not a form of de-
fault reasoning nor non-monotonic. It is
reasoning about one’s own knowledge or
belief. Hence it is called an auto-epistemic
reasoning.

Auto-epistemic reasoning models the reason-

ing of an ideally rational agent reflecting
upon his beliefs or knowledge.

Auto- Logics ate logics which describe the

reasoning of an ideally rational agent
reflecting upon his beliefs.

44
 Missionaries and Cannibals Revisited

McCarthy, 1985 revisits the problem: Three

missionaries and three cannibals come
to the river. A rowboat that seats two
is available. If the cannibals ever out-
number the missionaries on either bank
of the river, the missionaries will be
eaten. How shall they cross the river?

Traditionally the puzzler is expected to de-

vise a strategy of rowing the boat back and
forth that gets them all across and avoids
the disaster.

45
Traditional Solution: A state is a triple com-
prising the number of missionaries, canni-
bals and boats on the starting bank of the
river.

The initial state is 331, the desired state is

000,

A solution is given by the sequence: 331, 220,

321, 300,311, 110, 221, 020, 031, 010,
021, 000.

46
Imagine now giving someone a problem, and
after he puzzles for a while, he suggests
going upstream half a mile and crossing on
a bridge.

What a bridge? you say. No bridge is men-

tioned in the statement of the problem.

He replies: Well, they don’t say the isn’t a

bridge.

So you modify the problem to exclude the

bridges and pose it again.

He proposes a helicopter, and after you ex-

clude that, he proposes a winged horse
or that the others hang onto the out-
side of the boat while two row.

47
Finally, you tell him the solution.

He attacks your solution on the grounds that

the boat might have a leak or lack oars.

After you rectify that omission from the state-

ment of the problem, he suggests that a
see monster may swim up the river and
may swallow the boat.

Finally, you must look for a mode of rea-

soning that will settle his hash once and
for all.

48
McCarthy proposes circumscription as a
technique for solving his puzzle.

He argues that it is a part of common knowl-

edge that a boat can be used to cross
the river unless there is something with
it or something else prevents using it.

If our facts do not require that there be some-

thing that prevents crossing the river, cir-
cumscription will generate the conjecture
that there isn’t.

Lifschits has shown in 1987 that in some spe-

cial cases the circumscription is equivalent
to a first order sentence. item[In those
cases], we can go back to our secure and
well known classical logic.

49