A Report

on
Equivalence Class Testing

Submitted in partial fulfillment of

the requirement for the award of
the degree of
B.Tech in Artificial Intelligence and Data Science

Under The Supervision of

Dr. ALOK SIR
Professor

Submitted B
AMIR SOHAIL – 20scse1280004
ANKIT KUMAR - 20SCE1280003
AYUSHMAN RAOUT - 20SCSE1280031

SCHOOL OF COMPUTING SCIENCE AND ENGINEERING

DEPARTMENT OF COMPUTER SCIENCE AND
ENGINEERING GALGOTIAS UNIVERSITY, GREATER NOIDA
INDIA
DECEMBER, 2022
 ABSTRACT
Gödel’s two incompleteness theorems are among the most important results in modern
logic, and have deep implications for various issues. They concern the limits of provability
in formal axiomatic theories. The first incompleteness theorem states that in any consistent
formal system FF within which a certain amount of arithmetic can be carried out, there are
statements of the language of FF which can neither be proved nor disproved in FF.
According to the second incompleteness theorem, such a formal system cannot prove that
the system itself is consistent (assuming it is indeed consistent). These results have had a
great impact on the philosophy of mathematics and logic. There have been attempts to
apply the results also in other areas of philosophy such as the philosophy of mind, but
these attempted applications are more controversial. The present entry surveys the two
incompleteness theorems and various issues surrounding them. (See also the entry on Kurt
Gödel for a discussion of the incompleteness theorems that contextualizes them within a
broader discussion of his mathematical and philosophical work.)

Introduction
Gödel’s incompleteness theorems are among the most important results in modern logic.
These discoveries revolutionized the understanding of mathematics and logic, and had
dramatic implications for the philosophy of mathematics. There have also been attempts to
apply them in other fields of philosophy, but the legitimacy of many such applications is
much more controversial.
In order to understand Gödel’s theorems, one must first explain the key concepts essential
to it, such as “formal system”, “consistency”, and “completeness”. Roughly, a formal
system is a system of axioms equipped with rules of inference, which allow one to
generate new theorems. The set of axioms is required to be finite or at least decidable, i.e.,
there must be an algorithm (an effective method) which enables one to mechanically
decide whether a given statement is an axiom or not. If this condition is satisfied, the
theory is called “recursively axiomatizable”, or, simply, “axiomatizable”. The rules of
inference (of a formal system) are also effective operations, such that it can always be
mechanically decided whether one has a legitimate application of a rule of inference at
hand. Consequently, it is also possible to decide for any given finite sequence of formulas,
whether it constitutes a genuine derivation, or a proof, in the system—given the axioms
and the rules of inference of the system.
 The First Incompleteness Theorem—Proof Completed
To complete the proof, the Diagonalization Lemma is applied to the negated provability

(G)F⊢GF↔¬ProvF(┌GF┐).(G)F⊢GF↔¬ProvF(⌜GF⌝).
predicate ¬ProvF(x)¬ProvF(x): this gives a sentence GFGF such that

Thus, it can be shown, even inside FF, that GFGF is true if and only if it is not provable in FF.
It is not difficult to show that GFGF is neither provable nor disprovable in FF, if FF only is 1-consistent.

FF by ProvF(x)ProvF(x), FF would also prove ProvF(┌GF┐)ProvF(⌜GF⌝). However, because FF in fact

For the first half, assume that GFGF were provable. Then, by the weak representability of provability-in-

also proves the equivalence (G), i.e, F⊢GF↔¬ProvF(┌GF┐)F⊢GF↔¬ProvF(⌜GF⌝), FF would then

prove ¬GF¬GF too. But this would mean that FF is inconsistent. In sum, if FF is consistent, then GFGF is

For the second half, it has to be assumed that FF is 1-consistent (if ProvF(┌GF┐)ProvF(⌜GF⌝) has been
not provable in FF. For this first half, the assumption of the simple consistency of FF suffices.

chosen such that it is a Σ01Σ10-sentence; otherwise, the more general assumption of ωω-consistency is

Assume that F⊢¬GFF⊢¬GF. Then FF cannot prove GFGF, for otherwise FF would be simply
needed).

relation is strongly representable, for all nn, F⊢¬PrfF(n––,┌GF┐)F⊢¬PrfF(n_,⌜GF⌝). If

inconsistent. Hence no natural number nn is the Gödel number of a proof of GFGF, and because the proof

also F⊢∃xPrfF(x,┌GF┐)F⊢∃xPrfF(x,⌜GF⌝), FF is not 1-consistent, against the assumption.

Therefore FF does not prove ∃xPrfF(x,┌GF┐)∃xPrfF(x,⌜GF⌝), in other words, by the definition
of ProvF(x),FProvF(x),F does not prove ProvF(┌GF┐)ProvF(⌜GF⌝). By the key equivalence (G), FF also
does not prove ¬GF¬GF.
Gödel’s First Incompleteness Theorem
Assume FF is a formalized system which contains Robinson arithmetic Q. Then a sentence GFGF of the
language of FF can be mechanically constructed from FF such that:
If FF is consistent, then F⊬GFF⊬GF.
If FF is 1-consistent, then F⊬¬GFF⊬¬GF
Such an independent, or “undecidable” (that is, neither provable nor refutable
in F)F) statement GFGF in FF is often called “the Gödel sentence” of FF.
In fact, in favourable circumstances, it can be shown that GFGF is true, provided that FF is indeed
consistent. This is the case if, for example, the provability predicate ProvF(x)ProvF(x) has been chosen as

formula ∀x¬PrfF(x,┌GF┐)∀x¬PrfF(x,⌜GF⌝). Such formulas can be proved false whenever they in fact
a Σ01Σ10-formula: The Gödel sentence is then provably equivalent to the universal

are false: if false, there would be a number nn such that F⊢PrfF(n––,┌GF┐)F⊢PrfF(n_,⌜GF⌝) (this holds
already in Q). This, however, would contradict the incompleteness theorem. Therefore, GFGF cannot be
false, and must be true. For this reason, the Gödel sentence is often called “true but unprovable”.
One should not get confused here: “Gödel’s theorem” is the general incompleteness result of Gödel which
concerns a large class of formal systems, while the “Gödel sentence” is the constructed, formally
undecidable sentence which varies from one formal system to another. This is why it is important to
include the subscript FF in GFGF. Furthermore, one should not confuse the two different senses of
“undecidable” in this context. On the one hand, a particular sentence, like the Gödel sentence, may be
undecidable in the sense of being independent, i.e., neither provable not refutable in a chosen system. On
the other hand, a theory may be undecidable (see below) in the sense that there does not exist a decision
method for determining of an arbitrary given sentence of the language whether or not it is derivable in the
theory (so this latter sense of “undecidable” concerns, so to speak, an infinite class of statements).
In informal explanations of the first incompleteness theorem, it is often said that the Gödel
sentence GFGF “says of itself that it is not provable”. Such imprecise statements, however, should be
taken at least with a grain of salt. There are a number of reasons to conclude that, at least in general,
Gödel sentences do not really say anything substantial about themselves (Milne 2007 is a careful analysis
of such issues); for example, as was previously noted in the case of the Diagonalization Lemma, one is
usually operating here with mere material equivalences.
Rosser’s Improvement—From ωω-consistency to Consistency
In 1936, J. Barkley Rosser made an important improvement that allows one to get rid of the somewhat
clumsy assumption of ωω-consistency in the proof of Gödel’s first theorem. For this purpose, Rosser
introduced a new, somewhat artificial “provability predicate” Prov∗(x)Prov∗(x) which was constructed,
informally, as follows:
There exists yy such that yy is the Gödel number of a proof of the formula with Gödel number xx,
AND there does not exist zz smaller than yy such that zz is the Gödel number of a proof the negation of
the formula with Gödel number xx.
More formally:
Prov∗(x)=def∃y[PrfF(y,x)∧∀z<y(¬PrfF(z,neg(x)))],Prov∗(x)=def∃y[PrfF(y,x)∧∀z<y(¬PrfF(z,neg(x)))],
where PrfF(y,x)PrfF(y,x) is the more standard proof relation discussed earlier.
As it happens, if the formal system FF under consideration is indeed consistent, Rosser’s provability
predicate is co-extensional with the ordinary provability predicate. Applying the Diagonalization Lemma
to the negation of Rosser’s provability predicate Prov∗(x)Prov∗(x) gives:
Rosser’s modification of the first theorem (Rosser 1936)
Let FF be consistent formalized system which contains Q. Then there is a sentence RFRF of the language
of FF such that neither RFRF nor ¬RF¬RF is provable in FF.
 Incompleteness and Non-standard Models
It is illuminating to reflect on the first incompleteness theorem also from the model theoretic perspective
—though the theorem itself does not in any way require this. Namely, it is possible to conclude that any
theory FF satisfying the conditions of the theorem must possess, in addition to the intended interpretation
or “standard model” (in the case of arithmetical theories, the structure of natural numbers), non-intended
interpretations or “non-standard models”—that no such theory can rule out the latter and fix uniquely the
intended interpretation. Namely, if there are independent statements such as GF,FGF,F must have both

to ∃xPrfF(x,┌GF┐)∃xPrfF(x,⌜GF⌝), the latter models must possess entities which satisfy the
models which satisfy GFGF and models which rather satisfy ¬GF¬GF. As ¬GF¬GF is equivalent

formula PrfF(x,┌GF┐)PrfF(x,⌜GF⌝). And yet we know (because PrfF(x,y)PrfF(x,y) strongly represents

the proof relation) that for any numeral n––,Fn_,F can prove ¬PrfF(n––,┌GF┐)¬PrfF(n_,⌜GF⌝).
Therefore, no natural number nn can witness the formula. It follows that any such non-standard model
must contain, in addition to natural numbers (denotations of the numerals n––)n_), “infinite” non-natural
numbers after the natural numbers.
The study of non-standard models did not start with Gödel’s results—Skolem, in particular, was already
aware of them earlier in a different context (he had discovered that first-order theories of set theory have
unnaturally small, namely, countable models, in Skolem 1922; cf. the entry on Skolem’s paradox)—but
the first incompleteness theorem elucidates the existence of non-standard models in the context of
arithmetic, while the nonstandard models elucidate the first incompleteness theorem. Non-standard
models have since then become a rich research area in mathematical logic (see, e.g., Boolos & Jeffrey
1989: Ch. 17; Kaye 1991)

 The Second Incompleteness Theorem

 Preliminaries
Informally, the reasoning leading to the second incompleteness theorem is relatively simple. Given the
arithmetized provability predicate, it is also easy to present an arithmetized consistency statement: pick
some manifestly inconsistent formula (in arithmetical theories, a standard choice is (0–=1–)(0_=1_)); let

as ¬ProvF(┌⊥┐)¬ProvF(⌜⊥⌝). Let us abbreviate this formula by Cons(F)Cons(F). The proof of the first
us denote it by ⊥⊥; (the arithmetized counterpart of) the consistency of the system can then be defined

part of the first incompleteness theorem (i.e., the case (i) above) can then presumably be

F⊢Cons(F)→GF,F⊢Cons(F)→GF,
formalized inside FF (in practice this would certainly be intricate). This gives:

where GFGF is the Gödel sentence for FF provided by the first theorem. If Cons(F)Cons(F) were provable
in FF, so would be GFGF, by simple logic. This would contradict Gödel’s first theorem.
Consequently, Cons(F)Cons(F) cannot be provable in FF either.
Gödel’s second incompleteness theorem
Assume FF is a consistent formalized system which contains elementary arithmetic.
Then F⊬Cons(F)F⊬Cons(F).
There is a question of philosophical importance that should be mentioned here: As it stands, Gödel’s
second incompleteness theorem only establishes the unprovability of one sentence, Cons(F)Cons(F). But
does this sentence really express that FF is consistent? (Compare this with the remark above
that GFGF does not, strictly speaking, express its own unprovability.) Furthermore, might there not
be other sentences which are provable and also express the consistency of FF?
Giving a rigorous proof of the second theorem in a more general form that covers all such sentences,
however, has turned out to be very complicated. The basic reason for this is that, unlike in the first
theorem, not just any, merely extensionally adequate provability predicate works for the formalization of
the consistency claim. The manner of presentation makes all the difference. For example, Rosser’s
provability predicate mentioned above would not do; one can prove the “consistency” of FF in FF, if
consistency is expressed in terms of Rosser’s provability predicate. One must thus add some further
conditions for the provability predicate in order for the proof of the second incompleteness theorem to go
through. Following Feferman (1960), it is customary to say that whereas the first theorem and its relatives
are extensional results, the second theorem is intensional: it must be possible to think
that Cons(F)Cons(F) in some sense expresses the consistency of FF—that it really means that FF is
consistent.
 Derivability Conditions
The proof of the second incompleteness theorem requires that the provability predicate in FF satisfies a
number of conditions which are used in the details of the proof. There are several different sets of
conditions that will do.
The first detailed proof of the second incompleteness theorem appeared in (Hilbert & Bernays 1939)
(mainly written by Bernays), though only for one specific theory, PA. It uses a rather awkward set of
conditions for the provability predicate. These were more technical lemmas for the needs of a particular
proof and not any sort of analysis of “natural” provability predicates. A much more elegant, now standard
list of “derivability conditions” was presented by Löb (1955)—though their intended use was somewhat
different (see below).

(D3)F⊢A⇒F⊢ProvF(┌A┐).F⊢ProvF(┌A┐)→ProvF(┌ProvF(┌A┐)┐).F⊢ProvF(┌A┐)∧ProvF(┌A→
Löb’s Derivability Conditions(D1)(D2)

B┐)→ProvF(┌B┐).(D1)F⊢A⇒F⊢ProvF(⌜A⌝).(D2)F⊢ProvF(⌜A⌝)→ProvF(⌜ProvF(⌜A⌝)⌝).
(D3)F⊢ProvF(⌜A⌝)∧ProvF(⌜A→B⌝)→ProvF(⌜B⌝).
(D1) is simply a restatement of the requirement from the proof of the first theorem that provability is
weakly representable. Roughly put, (D2) requires that the whole demonstration of (D1), for the candidate
provability predicate ProvFProvF, can itself be formalized inside FF. Finally, (D3) requires that the
provability predicate is closed under Modus Ponens.
If the arithmetized provability predicate indeed satisfies these conditions, the second theorem can be
proved. Let GFGF once again be the Gödel sentence for FF given by the first theorem. It is not too

F⊢GF↔Cons(F).F⊢GF↔Cons(F).
difficult to show, using the derivability conditions, that:

This immediately yields the unprovability of Cons(F)Cons(F), given the first incompleteness theorem.
Furthermore, Jeroslow (1973) demonstrated, with an ingenious trick, that it is in fact possible to establish
the second theorem without (D3). However, in some other cases (e.g., when proving Löb’s theorem; see
below), and in Provability Logic, all three conditions are still needed.
 Feferman’s Alternative Approach to the Second Theorem
Under the assumption that a provability predicate for a theory satisfies the derivability conditions (or, by
Jeroslow’s trick, at least D1 and D2) it is relatively easy to prove the relevant case of the second
incompleteness theorem. However, in practise one has to establish whether a proposed arithmetized
provability predicate really satisfies the conditions case by case, and typically this is long and tedious.
This drawback, among other things (see Feferman 1997), led Solomon Feferman in the late 1950s to look
for an alternative line of attack to the second theorem (see Feferman 1960). Feferman approaches the
issue in two steps: First, he isolates the formulas ProvFOL(x)ProvFOL(x) which arithmetize some
standard notion of derivability in first-order logic in order to allow us to fix one chosen formula for
provability in logic. How the set of non-logical axioms of the system at issue are presented is left open at
this stage. Secondly, Feferman looks for a suitable constraint for presenting the axioms. Among the
formulas of the language of arithmetic, he isolates what he calls PR- and RE-formulas; the former
correspond to the canonical primitive recursive (PR) definitions in arithmetic, and the latter to existential
generalizations of the former. Every recursively enumerable (RE) set can be defined by a formula of the
latter sort; these are just the Σ01Σ10-formulas. These two classes are easy to discriminate purely by their
syntactical form. (In fact, by the MRDP Theorem (see below), one could—instead of RE-formulas—focus
on even simpler class of existentially quantified Diophantine equations.)
We have above noted the important fact that in all arithmetical theories FF containing Q, a set is strongly
representable in FF if and only if it is recursive, and a set is recursively enumerable if and only if it is
weakly representable. Furthermore, one can always take the formula weakly or strongly representing the
set to be a RE-formula (i.e., Σ01Σ10-formula; and, by MRDP Theorem, even an existentially quantified
Diophantine equation). It is then natural to require that the set of non-logical axioms of the system at issue
is represented by such a formula. If the arithmetized definition of the set of Gödel numbers of axioms
reflects how the axioms, if infinite, are inductively defined, the resulting formula will be Σ01Σ10. (For
theories which are axiomatizable with finitely many axioms, there is a unique representation of the axioms
in the form of a list, and consequently, a unique consistency statement relative
to ProvFOL(x)ProvFOL(x).) In contrast to determining whether the derivability conditions are satisfied, it
is a relatively routine task to determine that a given formula which formalizes the axioms is indeed of the
required form (Σ01Σ10).
Now the version of the second incompleteness theorem presented in Feferman 1960 is:
A variant of second incompleteness theorem (Feferman 1960)
Let FF be a consistent extension of PA, and let AxF(x)AxF(x) be a Σ01Σ10-formula which weakly
represents the axioms of FF, and Cons(F)Cons(F) be a consistency statement constructed
from AxF(x)AxF(x) and ProvFOL(x)ProvFOL(x). Then Cons(F)Cons(F) is not provable in FF.
For still different approaches to the second incompleteness theorem, see Feferman 1982, 1989a; Visser
2011. For some philosophical complications concerning the second theorem, see Detlefsen 1979, 1986,
1990, 2001; Auerbach 1985, 1992; Roeper 2003; Franks 2009 (see also section on incompleteness in the
entry on Hilbert’s program).
4. Results Related to the Incompleteness Theorems
 Tarski’s Theorem on the Undefinability of Truth
Gödel first arrived at the incompleteness results (see Section 5 below) by noting that truth (of the
language of a system) must be undefinable in the system, a result conventionally credited to Tarski (there
are certain real virtues in Tarski’s way of presenting the issue; see Gómez Torrente 2004). Let us now
view the result in the context of Tarski’s approach to truth.
Tarski clearly distinguished the object language, i.e., the language the truth of whose sentences is at stake,
and the metalanguage in which the former is discussed. He also required (see the entry on Tarski’s truth
definitions) that any satisfactory definition of truth True(x)True(x) for the object language should satisfy
his “Convention T”, that is, it should have as its consequence all equivalences (“T-equivalences”) of the

(T)True(┌A┐)↔B,(T)True(⌜A⌝)↔B,
form

where ┌A┐⌜A⌝ is a name of a sentence of the object language, and BB its translation in the
metalanguage. If the metalanguage is identical with the object language, or is an extension of the object

True(┌A┐)↔A.True(⌜A⌝)↔A.
language, BB is simply AA itself, and the T-equivalences are of the form:

What the undefinability theorem shows is that the object language and the metalanguage cannot coincide,
but must be distinct.
Tarski’s Undefinability Theorem
Let FF be a consistent formalized system that contains a sufficient amount of arithmetic. Then there is no

F⊢Tr(┌A┐)↔A.F⊢Tr(⌜A⌝)↔A.
formula Tr(x)Tr(x) in the language of FF such that for every sentence AA of the language of FF:

The idea of the proof: If there were such a formula of the language of FF, an easy application of the
Diagonalization Lemma to its negation would result in the paradoxical sentence LL (for “Liar”; see the

F⊢¬Tr(┌L┐)↔L,F⊢¬Tr(⌜L⌝)↔L,
Liar paradox)), such that:

which, together with the T-equivalences, which were assumed to be derivable, would quickly give an
explicit contradiction, thus contradicting the assumption that FF is consistent.
Similarly, it can be proved that the set of true sentences of FF is not definable in the intended
interpretation of FF—in the now standard sense of “definability” (see above).

REFERENCE-

 https://math.stackexchange.com/questions/63038/a-concrete-example-of-g%C3%B6dels-
incompleteness-theorem#:~:text=An%20example%20is%20the%20propositional,group%20do
%20not%20determine%20commutativity.

 https://plato.stanford.edu/entries/goedel-incompleteness/