New book:
Starting Science from God.
Links theism (religion) to science (psychology and physics) without
reduction.
Gödel's Results on the
Completeness
and Consistency of Mathematics
by Cameron C.
Pitcairn
Anyone who has had some contact with
mathematics sooner or later hears a rumor to the effect that back in the-1930's
a man named Kurt Gödel proved that there are statements in mathematics which
are in principle undecidable, i.e., statements which can neither be proved nor
disproved. Furthermore, he showed that there is no way of proving that
mathematics is consistent. Now I have deliberately stated these things in a
very rough form because this is the form in which most people have heard about
them. So much I had heard as a graduate student, but I never had an opportunity
in any of my courses to study_ these matters.
Of course, if you hear about
something like this and give it some thought, there are several natural
reactions. The first reactions: "How in the world did he prove something
like that?" The second reaction (which logically ought to be the first)
is: “What, exactly, did Gödel prove? What were his hypotheses and his
conclusions?” Further reflection on the proposition, "There are undecidable
statements in mathematics", entangles the casual reader in paradox. In my
own thinking, I can remember, the paradox took the following form.
Consider Goldbach’s
conjecture, which is the conjecture that any even integer greater than 2 is the
sum of two primes. For example, 4 = 2 + 2, 6 = 3 + 3, 8 = 5 + 3, 10 = 5 + 5,
12 = 7 + 5, and so on. You can keep on testing even integers until you get
tired, and then you can write a computer program to test several million for
you, and you will not find a counterexample to the conjecture. Naturally, a lot
of effort has been devoted to the search. for a proof of the conjecture, but
even though no one has found a counterexample, neither has anyone produced a
proof. And so the question arises: Goldbach's conjecture perhaps an undecidable
proposition? Supposing that it is in fact undecidable, what would that imply?
It would imply that 128, for example, is not a counterexample to the
conjecture, for if it were, that fact would be easily provable in a finite
number of steps; the proof would consist simply of finding the sum of every
pair of primes less than 128 and observing that no sum was equal to 128. Such a
proof would constitute a disproof of Goldbach's conjecture, contradicting our
assumption that the conjecture is undecidable. But the same argument applies
not only to 128 but also to any other definite integer as well. Hence, if Goldbach’
s conjecture is undecidable, then no integer provides a counterexample. But if
no integer provides a counterexample, then the conjecture must be true for
every integer, i.e., the conjecture s true. If Goldbach's conjecture is
undecidable, ten it is true! Is this possible? Can there be a statement which
is known to be true and yet which cannot be proved?
(Paradoxes like
this arise partly because our ordinary manner of. thinking is not precise
enough. Essentially what Gödel did was to apply a seemingly paradoxical
approach in such a precise way that he avoided all semantical fuzziness and
logical contradiction).
For some time,
then, I have casually wondered: What did Gödel prove? How did he prove it?
What implications or suggestions do his results have for mathematics or for
philosophy in general? Since I do not know much about the subject, I decided
that the best way to learn about it was to agree to give a talk about it, and thus
be forced to learn something about it. I therefore present myself to you today,
not as an expert in logic or the foundations of mathematics, but rather as a
fellow inquirer who may be able to share something of interest with you.
I will mention briefly the
mathematical history which led up to Gödel’s theorem. For a broader
perspective, I strongly recommend that you read Joel Pitcairn's 1956 address to
the Swedenborg Scientific Association. *1
Toward the end of the last century,
mathematicians seemed on the verge of establishing solid foundations for
mathematics in the theory of sets. This was a welcome development, because sets
are in some sense more transparent to our intuition than numbers and other
mathematical objects. Founding mathematics on the theory of sets, it was felt,
would
give mathematics a solid, certain
philosophical basis--or at any rate, as solid and certain as anyone could
reasonably expect. But then the para doxes started cropping up. Bertrand
Russell and others showed that the naive approach to set theory gave rise to
logical contradictions, and such inconsistency, of course, is fatal to any
putative system of mathematics. Ways were therefore sought to avoid the paradoxes
and at the same time to put mathematics on a secure footing.
One approach to the problem of
foundations was to reduce all of mathematics to the formal manipulation of
mathematical symbols, where this manipulation is conceived of as proceeding
according to a well defined set of rules without regard to any possible
meaning which might be given to those symbols. Although some of those
propounding this approach may have believed that mathematics "really
is" nothing but the systematic manipulation of meaningless symbols, such a
belief is by no means implicit in the approach itself. From one point of view,
this approach may be seen as an attempt to build a modal of mathematics
out of certain symbols and certain formal
rules governing those symbols. As with any model, there is no claim (or at
least, no necessary claim) that the model is identical with the- thing being
modeled, and, as with any model, the thing being modeled has certain features
which the model ignores. It is always the hope and intent of the model builder
that his model is adequate in spite of--or perhaps because of--the things which
it ignor.es. (It was Gödel’s accomplishment to show that no formal model of
mathematics can be entirely adequate.)
If a formal system of symbols is set
up, there has to be some way of talking about that system. If. we agree to call
such a system "mathematics", then the things we say about that system
are collectively called ''metamathematics". The distinction is between
statements in mathematics and statements about mathematics, and this is a very
necessary distinction if we are to see things clearly.
A formal system of mathematics can
be constructed from the following symbols. *2 In order to motivate the
introduction and use of these symbols, and make the exposition intuitively
easier to follow, I have given the ''meaning" of each symbol, but you must
r ember that the for mal rules governing the system make no reference whatever
to the meaning of the symbols, and, strictly speaking, neither should we; the
system is one from which all meaning has been abstracted.
Symbol Meaning Symbol
Meaning
~ not 0 zero
V or s successor
of (an integer)
→ implies (
€ there exists ) punctuation
= equals ,
In addition to the above symbols,
there must be symbols to serve as names of variables. A countable infinity of
these will do nicely.
In addition to a set of symbols, our
formal system requires a set of rules to tell us which collections of symbols
'1make sense" and which do not. Our rules might tell us, for example, that
"O = O" is a collection of symbols which "makes sense",
whereas "= = (" is not. Those collections of symbols which conform to
these rules are called formulas of the formal system; they correspond to
statements in everyday language which are grammatically correct (but not
necessarily true). In this paper I will use interchangeably the phrases
"formula", "mathematical statement", and "statement
within formal mathematics".
I will not give in detail the rules
for constructing formulas, but content myself with the following single example
of such a rule: If a given collection of symbols is a formula, then the
collection of symbols obtained by writing "~(", followed by the given
collection, followed
by ")", is also a formula. (If,
for example, we know that "0=0" is a formula, the rule just
enunciated allows us to conclude that "~(0=0)" is a formula. In
other words, the negation of a formula is also a formula.)
Notice that our system does not include
any numbers other than O. However, the successor function enables us to define
all of the natural numbers. For example, “1” is defined to represent
"sO", "2" is defined to represent "s1", which in
turn represents "ssO", and so on.
From the formal point of view, a symbol
like 11211 merely represents an abbreviation for the set of symbols "ssO",
and any formula which contains defined symbols like "2" can in
principle be rewritten to include only primitive symbols like "s" and
"0".
It turns out that not only the
natural numbers but all the ordinary mathematical concepts--such as addition,
multiplication, prime numbers, and so on--can be defined (ultimately) in terms
of our ten primitive symbols, together with our countable number of variables.
To finish our formal system, we must
include a mechanism for distinguishing the "true" statements from the
others. We can start with certain specified statements, called
"axioms", and give a set of mechanical rules for deriving sequences
of formulas, where each formula is either
an axiom, or else depends (by some
definite rule of symbol manipulation) on the formulas which precede it in the
sequence. Such a sequence of formulas is called a "proof", and in
fact is a model of what we intuitively mean by the word "proof". Each
statement follows from the statements before it in some well-defined way. Since
the "rules of inference" are strictly mechanical, we can talk about a
"proof" in this system without any recourse to psychological
certainty or other such vague and debatable concepts.
The last statement of a proof is, of
course, called "theorem" (or, equivalently, a "provable
statement") of the system. If I assert that a certain formula of the
system is provable, I mean that there exists a proof in the system which has
that formula as its last formula. Although the question of whether given
formula is or is not provable may be problematical, once a proof has been
discovered it is easy to verify that the proof is correct simply by following the
rules of inference.
In 1904,
David Hilbert propounded the following goal for mathematicians: first to show
that all of known mathematics could be mirrored in a formal system, and then to
how, by finite constructive methods of reasoning applied to the formal system,
that the system is consistent. A system is said to be consistent if there is no
theorem of the system whose negation is also a theorem of the system. For
example, in a consistent system it is not possible to prove both
"0=0" and "~ (0=0)". In fact, to say that the system is inconsistent
is equivalent to saying that every statement in the system is provable. *3
Principia Mathematica, by
Bertrand Russell and Alfred North Whitehead, set forth a formal system, of the
sort envisioned by Hilbert, which · attempted to account for all of
mathematics. In the 19201 s, other results encouraged mathematicians to believe
that Hilbert's goal was very close to attainment. A proof of the consistency of
formal mathematics, by finite constructive methods, seemed "just around
the corner." Then in 1931 came the bombshell: Gödel proved that if the
formal system embodied in Principia Mathematica is consistent, then its
consistency cannot be proved by any method which is acceptable as a method of
proof within Principia Mathematica. Gödel’s proof, moreover, applies not
only to Principia Mathematica, but to any formal system which is
sufficiently powerful to include the natural numbers (and a fortiori to any
system which claims to encompass all of mathematics):- This disheartening result
is an almost immediate consequence of the fact, which Gödel proved in the same
paper, that any such formal system is incomplete, in the sense that if the
system is consistent, then there exists a statement in the system which is not
provable within the system and whose negation is not provable within the
system. (Such a statement is called undecidable.) Gödel proved that undecidable
statements exist by constructing an example of one. *4
In the next few minutes, I want to give
you some idea of how Gödel went about constructing his example of an
undecidable statement, although his construction was more involved and delicate
than I have time to indicate here.
One of the ancient paradoxes is the
paradox of the liar. A man says, "I am telling a lie right now." Is
he telling the truth or not? Another paradox of more recent vintage, propounded
by Jules Richard, is as follows: Consider the set of all predicates in the
English language which describe properties of integers: for example, "is
prime", "has two prime factors", "is greater than fifteen
and less than twenty-one", and so on. Since each such predicate consists
of a finite number of letters, we can arrange all such predicates in some
definite order--for example in order of increasing length, and alphabetically among
predicates of the same length. Such an ordering gives rise to a numbering of
the predicates, with the first predicate in the list bearing the number 1, the
second bearing the number 2, and so on. We now define a number n to be Richardian
if and only if n does not satisfy the predicate whose number is n. (For example,
"is prime" may be predicate number 1000 on our list; if this is the
case, then 1000 is Richardian, because "1000 is prime" is false.)
Consider now the predicate "is Richardian", and suppose it to be
predicate number r on our list. Is r Riehardian? If r is Richardian, then r
satisfies the predicate "is Richardian", and hence r is not
Richardian. Conversely, if r is not Richardian then r is Richardian.
An essential feature of Richard's
paradox, as of the liar's paradox, is that it involves statements which
(directly or indirectly) refer to themselves.
Recall that Gödel’s Incompleteness
Theorem says that any sufficiently rich consistent formal system contains an
undecidable statement. Essentially, the way Gödel proved this was to construct
a statement which says of itself that it is not provable. Now, strictly
speaking it is not possible for a statement within mathematics to refer to
itself, since such statements are not about statements but about numbers (and
other mathematical objects). Statements about statements belong to metamathematics.
Gödel surmounted this obstacle by demonstrating a correspondence whereby
certain statements in metamathematics (which are statements about
mathematical statements) can be translated into statements in mathematics
(which are statements about numbers). He showed that for each formula
and for each sequence of formulas in formal mathematics, we can compute a
corresponding integer (now called the Gödel number of the formula or the
sequence), and this is done in such a way that if we are given the Gödel number
of any formula (or sequence of formulas), we can, by factoring the number into
primes, compute the original formula (or sequence of formulas).
How are these Gödel numbers used? Gödel
showed how to translate certain metamathematical statements, which are
statements about formulas, into mathematical statements which assert that the Gödel
numbers of those formulas belong to certain sets, and this translation is done
in such a way that provable mathematical statements represent true metamathematical
statements.
In particular, Gödel
constructed a mathematical statement, G, which corresponds to the metamathematical
statement "G is not provable." Now if G is provable, then the
metamathematical statement, "G is not provable", is true, which is a
contradiction. If ~G is provable, then the metamathematical statement, "G
is provable", is true, but we have just seen that G cannot be provable. Hence
neither G nor ~G is provable, i.e., G is undecidable.
Note, by the way, that G, though not
provable, is nonetheless true. This may sound like a contradiction; but recall
that "provable" here means "formally provable within the system
by means of a given set of axioms and rules of inference." The
metamathematical reasoning
which tells us that G is true does not
constitute a "proof" in this sense, nor can it be mirrored by any
proof within the system.
Let us now carry the argument a step
further. Let us call the formal system under discussion "S". In our
previous remarks we have sketched a proof of the metamathematical statement:
"If S is consistent, then G is not provable." Gödel showed that this
statement can be represented by a statement H within the system S. H, in other
words, is equivalent to "C implies G", where C is the statement
"S is consistent." H is, in fact, provable in the system S. Now
suppose that C is provable in S. The rules of inference allow us to combine
the provable statements C and H to arrive at a proof of G. We already know,
however, that if the system S is consistent then G is not provable. We conclude
that if the system S is consistent, then C is not provable in S. In other
words, no proof that S is consistent can be mirrored by a proof within S. This
is Gödel’s famous result on the consistency of mathematics.
And so what? I would certainly
hesitate to say that a result like this proves something about philosophy, but
it may suggest a lot of interesting ideas. The remarkable thing about
this proof is that it beats the formalists at their own game. The formalists
proclaim that mathematics is nothing but a meaningless collection of symbols,
but Gödel showed that there is always something beyond: there are statements
which we know are true and which are not accounted for by the formal system,
and this holds no matter how much we enlarge the system. In a sense, Gödel
showed that there is a higher degree of mathematics which you cannot achieve by
manipulating symbols. His methods, moreover, were unimpeachable from any point
of view.
Gödel himself appeared to be a
thoroughgoing Platonist. He wrote: "Classes and concepts may … be
conceived as real objects …. It seems to me that the assumption of such objects
is quite as legitimate as the assumption of physical bodies and there is quite
as much reason to believe in their existence. They are in the same sense
necessary to obtain a satisfactory system of mathematics as physical bodies are
necessary for a satisfactory theory of our sense perceptions, and in both cases
it is impossible to interpret the propositions one wants to assert about these
entities as propositions about the “data”, i.e., in the latter case the
actually occurring sense perceptions. *5
Joel Pitcairn suggests an analogy
with the Heisenberg uncertainty principle; in the one case we have a trade-off
between position and momentum, and in the other case we have a trade-off
between completeness and consistency. *6 (You can have a consistent system, or
you can have
a complete system, but you cannot have
both.)
Starting another line of thought, we
might wonder whether it will ever be possible to construct a computer to
perform the functions of a mathematician (whatever they are!). One way of
interpreting Gödel’s results is to say that any system which consists simply of
the manipulation of symbols according to a given set of rules is incapable of
taking in all of mathematics. Since a computer is, after all, essentially a
mechanical realization of just such a system, we might plausibly conclude that
no computer is capable of replacing the creative mathematician. This might also
suggest that the way in which the human mind works is in some respects very
different from the way in which a computer works.
In the eighteenth century, David
Hume demonstrated in a fairly conclusive way that the empirical method of
science cannot be rationally justified simply on the basis of observation. The
empirical method depends fundamentally on the inductive principle, i.e., the
principle that the world will behave in the future as it has been observed to
behave in the past. Since no amount of observation can confirm the inductive
principle, unless we assume the principle in some form to begin with, it follows
that to believe the principle at all is an act of faith. There is no way by
which one can pull himself by his bootstraps up from the level of sensory data
to the discretely higher level of empirical reasoning. Similarly, to believe in
the consistency, and hence the meaning, of mathematics requires an act of
faith. "Suppose," wrote Frank DeSua, "we loosely define a
religion as any discipline whose foundations rest on an element of faith,
irrespective of any element of reason which may be present. Quantum mechanics
for example would be a religion under this definition. But mathematics would
hold the unique position of being the only branch of theology possessing a rigorous
demonstration of the fact that it should be so classified." *7
I want to conclude this talk by
suggesting the applicability of the doctrine of discrete degrees. Joel Pitcairn
has pointed out that many of the confusions which plague modern philosophy stem
from a failure to take into account the. separation which exists between that
which thinks and that which is thought about, and he asserts that this "is
an instance of the confusion of discrete degrees, and in our opinion no truly
adequate philosophy of science can be possible which does not depend explicitly
on the doctrine of discrete degrees." *8
This may seem like a rather sweeping
statement, but it is justified by the following from Divine Love and Wisdom:
"A knowledge of degrees is like a key for uncovering and penetrating into
the causes of things.
Without this knowledge hardly anything of
cause can be known; for, without it, objects and subjects of both worlds seem
so simple, as if there were nothing in them beyond that which meets the eye;
when yet, compared to the things which lie hidden within, what is thus seen is
as one to thousands, nay rather to tens of thousands. The interiors which do
not lie open can in no wise be laid bare except by a knowledge of degrees; for
exteriors move towards interiors and through these towards inmosts, by degrees;
not by continuous, but by discrete degrees." *9
I would suggest that what I have been
talking about today provides us with an example of discrete degrees,
metamathematics being a discrete degree above formal mathematics--or perhaps it
would be more accurate to say that the meaning of mathematics is a discrete
degree above the symbols of mathematics. Gödel achieved his results by
separating the two levels--avoiding confusion between them--but then relating
them to each other and exploiting the connection. The doctrine of discrete
degrees is a large one, and I do not have the time here to go into it deeply,
but it seems to me to throw a very powerful light on our subject, and our
subject may in turn serve as a confirmation of the doctrine. *10
Closely connected with' the idea of
discrete degrees is the idea of discrete levels of the human mind. Rational
Psychology refers to the "pure intellect", which is that faculty
of the mind which grasps truth whole, the faculty which looks at a thing and
knows immediately (i.e., without mediation) that it is true. It is a faculty
discretely higher than the faculty which allows us to manipulate symbols; the
pure intellect is, I suggest, the faculty which invests those symbols with
meaning. I cannot help feeling that Gödel’s results lend powerful confirmation
to the existence—I might almost say, the logical necessity- of such a faculty.
*11
"Those whose
thought or rational analysis more closely approaches this pure intellect have
an instantaneous sight and recognition of many propositions as being true or
false, and this without posteriori demonstration from effects, experience,
artificial logic, and the scholas tic sciences. Frequently, indeed, this is so
much the case that they are indignant that the mind should wish to demonstrate
things which in themselves, are clearer, more sure, truer, and higher than any
demonstration. Attempts at such demonstrations they consider as so many
twilight shadows which do not enlighten but rather obscure. Of such sort are we
when we become pure intelligences or souls, for then we shall laugh at literary
sports as the sports of infants, and at the whole syllogistic logic as a
child's game of even and odd." *12
A good note on
which to end, I think.
August
14, 1979
NOTES
1. Joel Pitcairn, "The
Foundations of Mathematics," The New Philosophy, LIX (July, 1956), PP•
79-91.
2. The particular formal system
presented here is taken from Ernest Nagel and James R. Newman, Gödel’s Proof
(New York: New York University Press, 1958).
3. The principle that in an
inconsistent system every statement is provable is an easy consequence of
elementary logic. According to my recollection, this principle was once
enunciated in a popular lecture by Bertrand Russell as: "If I assume the
truth of just one false statement, I can prove anything." Someone in the
audience called out, "Two plus two equals three. Prove you are the Pope.11 Russell
replied, "If two plus two equals three, then, by subtracting two from both
sides of the equation, we have two equals
one. The set consisting of Bertrand Russell and the Pope contains just two
elements, that is
to say (since two equals one), it contains
just one element. Since Russell and the Pope are elements of a set containing-
just one element, they are identical; therefore, I am the Pope!"
4. Gödel’s paper, translated into
English, is reprinted in Jean van Heijenoort, ed., From Frege to Gödel: A Source
Book in Mathematical Logic, 1879-1931 ( Cambridge, Y ss.: Harvard University
Press, 1967), pp. 592-617. A leisurely exposition for the layman may be found
in Nagel and Newman, op. cit. A brief but incisive sketch of the proof is given
in Nic las Bourbaki, Theory of Sets (Reading, Mass.: Addison-Wesley, 1968), pp.
339-341. - --
I should note here that Gentzen
subsequently gave a proof of the consistency of arithmetic. This does not'
contradict Gödel’s result, since Gentzen used a method of proof, namely
transfinite
induction, which is not allowable within
the formal system which he was
considering. Roughly speaking, we can
interpret Gödel’s result
as saying that any proof of the
consistency of mathematics must use methods which are of a ''higher order"
than any methods of proof which are allowed within mathematics, and of course
the validity of these "higher order" methods will be at least as suspect
as the validity
of mathematics itself, so that such a
proof doe s little if anything to
put the foundations on a firmer footing.
5. Kurt Gödel, "Russell's
Mathematical Logic," in Philosophy of Mathematics: Selected Readings, ed.
by Paul Benacerraf and Hilary Putnam (Englewood Cliffs, N.J.: Prentice-Hall,
1964), p. 220.
6. Joel Pitcairn, op. cit. p. 89.
7. Frank DeSua, "Consistency
and Completeness - - A Résumé," The American Mathematical Monthly,
LXIII ( May, 1956),p. 305.
8. Joel Pitcairn, review of The
Scientific A venture by Herbert Dingle, The New Philosophy, LVII (October,
1954), p. 257.
9. Divine Love and
Wisdom, number 184.
10. For more
concerning the doctrine of discrete degrees, see Divine Love and Wisdom, number
256, Arcana Coelestia numbers 3691 and 5114, and many other passages,
especially in Divine Love and Wisdom.
11. Cf. Joel
Pitcairn, “The Foundations of Mathematics,” op. cit. pp. 80-81.