Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I am trying very hard to understand Gödel's Incompleteness Theorem. I am really interested in what it says about axiomatic languages, but I have some questions:

Gödel's theorem is proved based on Arithmetic and its four operators: is all mathematics derived from these four operators (×, +, -, ÷) ?

Operations such as log(x) and sin(x) are indeed atomic operations, like those of arithmetic, but aren't there infinitely many such operators that have inverses (that is, + and - are "inverse" operations, × and ÷ are inverse operations).

To me it seems as though making a statement about the limitations of provability given 4 arbitrary operators is absurd, but that probably highlights a gap in my understanding, given that he proved this in 1931 and its unlikely that I have found a counter-argument.

As a follow-up remark, why the obsession with arithmetic operators? They probably seem "fundamental" to us as humans, but to me they all seem to be derived from four possible graphical arrangements of numbers (if we consider four sides to a digit), and fundamentally derived from addition.

[][] o [][] addition and, its inverse, subtraction

[][]
[][] multiplication (iterative addition) and, its inverse, division

There must be operators that are consistent on the natural numbers that we certainly aren't aware of, no?

Please excuse my ignorance, I am hoping I haven't offended any real mathematicians with this posting.


edit: I think I am understanding this a lot more, and I think my main difficulty in understanding this was that:

There are statements that are true that are unprovable.

Seemed like an impossible statement. It does, however, make sense to me at the moment in the context of an axiomatic language with a limited number of axioms. Ultimately, suggesting that there are statements that are true and expressible in the language, but are unprovable in the language (because of the limited set of axioms), is what I believe to be the point of the proof -- is this correct?

share|improve this question
    
It might be expeditious to share what book or other material you are studying and at what level. Gödel actually has two "incompleteness" theorems. The subject matter is formal first-order theories, and in particular a (partial) axiomatization of the theory of natural numbers. Certainly the operations of addition and multiplication enter into the proof, but subtraction and division are not defined in generality for natural numbers. So it is not a matter of basing a proof on "4 arbitrary operators" nor of concluding anything inconsistent about operators. –  hardmath Jan 4 '11 at 21:38
    
@hardmath: No book, just internet; recommendations welcome. I come from a CS background and have had my fair share of math, but nothing remotely theoretical. I don't know much about number theory or linear algebra, but I am also a linguist and have studied grammars and their generation (in the CS / Chomsky sense of the word). –  sova Jan 4 '11 at 22:02
3  
It is important not to think of this result as coming from axioms that are too weak. You can add an axiom to make the specific sentence provable, but then another sentence becomes a witness. –  Ross Millikan Jan 4 '11 at 23:03
    
The incompleteness theorem says that any reasonable (i.e. consistent and axiomatizable) extension (by any new function/relation symbols and axioms) of the weak theory about arithmetic is incomplete. Using a weaker base theory in the theorems is a stronger result since it means that more theories are incomplete. –  Kaveh Feb 28 '11 at 1:30
    
@Kaveh: thank you for that insight, I've done a bit of reading on the topic since I posted this question and your answer makes sense on first read. –  sova Feb 28 '11 at 2:33

5 Answers 5

up vote 65 down vote accepted

There's a fair amount of historical context to make some of Gödel's choices in his original proof clear. For a good overview of the proof, Nagel and Newman's Goedel's Proof is pretty good (though not without its detractors). I also highly recommend the late Torkel Franzen's Godel's Theorem: An Incomplete Guide to its Use and Abuse. I may myself be guilty of some of those abuses below; I'm trying to shy away from a lot of the technical details, and this usually invites imprecision and even abuse in this particular field. I hope those who know better than me will keep me honest via comments and appropriate ear-pulling.

Some history. Sometimes I think of the 19th century as the Menagerie century. A "menagerie" was like a zoo of weird animals. During a lot of the 19th century, people were trying to clean up some of the logical problems that abounded in the foundations of mathematics. Calculus plainly worked, but a lot of the notions of infinitesimals were simply incompatible with some 'facts' about real numbers (eventually solved by Weierstrass's notion of limits using $\epsilon$s and $\delta$s). A lot of assumptions people had been making implicitly (or explicitly) were shown to be false through the construction of explicit counterexamples (the "weird animals" that lead me to call it the menagerie century; many mathematicians were like explorers bringing back weird animals nobody had seen before and which challenged people's notions of what was and was not the case): the Dirichlet function to show that you can have functions that are discontinuous everywhere; functions that are continuous everywhere but nowhere differentiable; the failure of Dirichlet's principle for some functions; Peano's curve that filled-up a square; etc. Then, the antinomies (paradoxes and contradictions) in the early set theory. Even some work which today we find completely without problem caused a lot of debate: Hilbert's solution of the problem of a finite basis for invariants in any number of variables was originally derided by Gordan as "theology, not mathematics" (the solution was not constructive and involved the use of an argument by contradiction). Many found a lot of these developments deeply troubling. A foundational crisis arose (see the link in Qiaochu's answer).

Hilbert, one of the leading mathematicians of the late 19th century, proposed a way to settle the differences between the two main camps. His proposal was essentially to try to use methods that both camps found unassailably valid to show that mathematics was consistent: that it was not possible to prove both a proposition and its negation. In fact, his proposal was to use methods that both camps found unassailable to prove that the methods that one camp found troublesome would not introduce any problems. This was the essence of the Hilbert Programme.

In order to be able to accomplish this, however, one needed to have some way to study proofs and mathematics itself. There arose the notion of "formal proof", the idea of an axiomatization of basic mathematics, etc. There were several competing axiomatic systems for the basics of mathematics: Zermelo attempted to axiomatize set theory (later expanded to Zermelo-Fraenkel set theory); Peano had proposed a collection of axioms for basic arithmetic; and famously Russell and Whitehead had, in their massive book Principia Mathematica attempted to establish an axiomatic and deductive system for all of mathematics (as I recall, it takes hundreds of pages to finally get to $1+1=2$). Some early successes were achieved, with people showing that some parts of such theories were in fact consistent (more on this later). Then came Gödel's work.

Consistency and Completeness. We say a formal theory is consistent if you cannot prove both $P$ and $\neg P$ in the theory for some sentence $P$. In fact, because from $P$ and $\neg P$ you can prove anything using classical logic, it is equivalent that a theory is consistent if and only if there is at least one sentence $Q$ such that there is no proof of $Q$ in the theory. By contrast, a theory is said to be complete if given any sentence $P$, either the has a proof of $P$ or a proof of $\neg P$. (Note that an inconsistent theory is necessarily complete). Hilbert proposed to find a consistent and complete axiomatization of arithmetic, together with a proof (using only the basic mathematics that both camps agreed on) that it was both complete and consistent, and that it would remain so even if some of the tools that his camp used (which the other found unpalatable and doubtful) were used with it.

Why arithmetic? Arithmetic was a particularly good field to focus on in the early efforts. First, it was the basis of the "other camp". Kronecker famously said "God gave us the natural numbers, the rest is the work of man." It was hoped that an axiomatization of the natural numbers and their basic operations (addition, multiplication) and relations (order) would be both relatively easy, and also have a hope of being both consistent a complete. That is, it was a good testing ground, because it contained a lot of interesting and nontrivial mathematics, and yet seemed to be reasonably simple.

Gödel. Gödel focused on arithmetic for this reason. As it happens, multiplication is key to the argument (there is something special about multiplication; some theories of the natural numbers that include only addition can be shown to be consistent using only the kinds of tools that Hilbert allowed). To answer one of your questions along the way, Gödel even defined new operations and relations on natural numbers that had little to do with addition and multiplication along the way, so that yes, there are operations other than those (no fetishism about them at all). But in fact, Gödel did not restrict himself solely to arithmetic. His proof is, on its face, about the entire system of mathematics set forth in Russell and Whitehead's Principia, though as Gödel notes it can easily be adapted to other systems so long as they satisfy certain criteria (that's why Gödel's original paper has a title that explicitly refers to the Principia "and related systems").

What Gödel showed was that any theory, subject to some technical restrictions (for example, you must have a way of recognizing whether a given sentence is or is not an axiom), that is "strong enough" that you can use it to define a certain portion of arithmetic will necessarily be either incomplete or inconsistent (that is, either you can prove everything in that theory, or else there is at least one sentence $P$ such that neither $P$ nor $\neg P$ can be proven). It's not a limitation based on four operations, or an obsession with those operations: quite the opposite. What it says if that if you want your theory to include at least some arithmetic, then your theory is going to be so complicated that either it is inconsistent, or else there are propositions that can neither be proven nor disproven using only the methods that both camps found valid.

That is: what it shows is a limitation of those particular (logically unassailable) methods. If we use other methods, we are able to establish consistency of arithmetic, for example, but if you had your doubts about the consistency of arithmetic in the first place, chances are you will find those methods just as doubtful.

Now, about you coda, and the statement "statements that are true but unprovable"; this is not very apt. You will find a lot of criticism to this paraphrase in Franzen's book, with good reason. It's best to think that you have statements that are neither provable nor disprovable. In fact, one of the things we know is that if you have such a statement $P$, in a theory $M$, then you can find a model (an interpretation of the axioms of $M$ that makes the axioms true) in which $P$ is true, and a different model in which $P$ is false. So in a sense, $P$ is neither "true" nor "false", because whether it is true or false will depend on the model you are using. For example, the Gödel sentence $G$ that proves the First Incompleteness Theorem (that if arithmetic is consistent then it is incomplete, since there can be no proof of the sentence $G$ and no proof of $\neg G$) is often said to be "true but unprovable" because $G$ can be interpreted as saying "There is no proof of $G$ in this theory." But in fact, you can find a model of arithmetic in which $G$ is false, so why do we say $G$ is "true"? Well, the point is that $G$ is true in what is called "the standard model." There is a particular interpretation of arithmetic (of what "natural number" means, of what $0$ means, of what "successor of $n$" means, of what $+$ means, etc) which we usually have in mind; in that model, $G$ is true but not provable. But we know that there are different models (where 'natural number' may mean something completely different, or perhaps $+$ means something different) where we can show that $G$ is false under that interpretation of the axioms. I would stay away from "true" and "false", and stick with "provable" and "unprovable" when discussing this; it tends to prevent problems.

First Summary. So: there were historical reasons why Gödel focused on arithmetic; the limitation is not of arithmetic itself, but rather of the formal methods in question: if your theory is sufficiently "strong" that it can represent part of arithmetic (plus it satisfies the few technical restrictions), then either your theory is inconsistent, or else the finitistic proof methods at issue cannot suffice to settle all questions (there are sentences $P$ which can neither be proven nor disproven).

Can something be rescued? Well, ideally of course we would have liked a theory that was complete and consistent, and that we could show is complete and consistent using only the kinds of logical methods which we, and pretty much everyone else, finds beyond doubt. But perhaps we can at least show that the other methods don't introduce any problems? That is, that the theory is consistent, even if it is not complete? That at least would be somewhat of a victory.

Unfortunately, Gödel also proved that this is not the case. He showed that if your theory is sufficiently complex that it can represent the part of arithmetic at issue (and it satisfies the technical conditions alluded to earlier), and the theory is consistent, then in fact one of the things that it cannot settle is whether the theory is consistent! That is, one can write down a sentence $C$ which makes sense in the theory, and which essentially "means" "This theory is consistent" (much like the Gödel sentence $G$ essentially "means" that "there is not proof of $G$ in this theory"), and which one can prove that if the theory is consistent then the theory has no proof of $C$ and no proof of $\neg C$.

Again, this is a limitation of those finitistic methods that everyone finds logically unassailable. In fact, there are proofs of the consistency of arithmetic using transfinite induction, but as I alluded to above, if you harbored doubts about arithmetic in the first place, you are simply not going to be very comfortable with transfinite induction either! Imagine that you are not sure that $X$ is being truthful, and $X$ suggests that you ask $Y$ about $X$s truthfulness; you don't know $Y$, but $X$ assures you that $Y$ is very trustworthy. Well, that's not going to help you, right?

Key take-away: Because the theorem applies to any theory that is sufficiently complex (and satisfies the technical restrictions), we are not even in a position of enlarging our set of axioms to escape these problems. So long as we restrict ourselves to enlargement methods that we find logically unassailable, the technical restrictions will still be satisfied, so that the new theory, stronger and larger though it will be because it has more axioms, will still be incomplete (though it will possibly be other sentences that are now incomplete or unprovable; remember that by adding axioms, we are also potentially expanding the kinds of things about which we can talk). So the theorems are not about shortcomings of particular axiomatic systems, but rather about those finitistic methods within a very large class of systems.

What about those 'technical restrictions'? They are important. Suppose that arithmetic were consistent. That means that there is at least one model for it. We could pick a model $M$, and then say "Let's make a theory whose axioms are exactly those sentences that are true when interpreted in $M$." This is a complete and consistent axiomatic system for arithmetic. Complete, because each sentence is either true in $M$ (and hence an axiom, hence provable in this theory) or else false in $M$, in which case its negation is true (and hence an axiom, and hence provable). And consistent, because it has a model, $M$, and a theory is consistent if and only if it has a model. The problem with this axiomatic theory is that if I give you a sentence, you're going to have a hard time deciding if it is or it is not an axiom! We didn't really achieve anything by taking this axiomatic system. The "technical restrictions" are both in the form of making the system actually usable, and also certain technical issues that arise from the mechanics of the proof. But the restrictions are mild enough that pretty much everyone agrees that most reasonable theories will likely satisfy them.

Second summary. So: if you have a formal axiomatic system which satisfies certain technical (but mild) restrictions, if the theory is large enough that you can represent (a certain part of) arithmetic in it, then the theory is either inconsistent, or else the finitistic methods that everyone agrees are unassailable are insufficient to prove or disprove every sentence in the theory; worse, one of the things that the finitistic methods cannot prove or disprove is whether the theory is in fact consistent or not.

Hope that helps. I really recommend Franzen's book in any case. It will lead you away from potential misinterpretations of what the theorem says (I am likely guilty of a few myself above, which will no doubt be addresssed in comments by those who know better than I do).

share|improve this answer
1  
@Arturo: Do you know where to read some of the "detractors" you mention at the beginning? I would like to see what objections they could possibly have. –  Andres Caicedo Jan 5 '11 at 2:58
1  
About Gordan's comment: cs.nyu.edu/pipermail/fom/1999-August/003297.html –  Andres Caicedo Jan 5 '11 at 3:01
1  
(This is an excellent answer, by the way!) –  Andres Caicedo Jan 5 '11 at 3:02
    
@Andres: I'm away from home right now, so I can't check, but I have the distinct impression that Torkel Franzen had some criticism on the presentation in Nagel and Newman (the informal part, at least). In sci.math some time ago (groups.google.com/group/sci.math/msg/f98ba1ee8212eeaa), Aatu Koskensilta made passing remarks about it being "marred with all sorts of misconceptions", though he did not elaborate. I've seen that kind of comment on and off, from time to time. –  Arturo Magidin Jan 5 '11 at 3:05
1  
@Andres: And thanks for the kind words! –  Arturo Magidin Jan 5 '11 at 3:22

The incompleteness theorem is much more general than "arithmetic and its four operators." What it says is that any effectively generated formal system which is sufficiently powerful is either inconsistent or incomplete. This requires that the system be capable of talking about arithmetic, but it can also talk about much more than arithmetic.

To me it seems as though making a statement about the limitations of provability given 4 arbitrary operators is absurd

Then you should study harder! Mathematics is not beholden to your intuition. When your intuition clashes with the mathematics, you should assume that your intuition is wrong. (I am not trying to be harsh here, but this is an attitude you need to digest before you can really learn any mathematics.)

You also shouldn't think of arithmetic as just being about the arithmetic operations: it's also about the quantifiers.

As a follow-up remark, why the obsession with arithmetic operators?

Who claims to have an obsession with arithmetic operators? Perhaps what you are missing here is historical context. The historical context is, roughly, this: back in Gödel's day there was a program, initiated by Hilbert, which sought to give a complete set of axioms for all of mathematics. That is, Hilbert wanted to write down a set of (consistent) axioms from which all mathematical truths were deducible. (To really understand why Hilbert wanted to do this you should read about the foundational crisis in mathematics). This was very grand and ambitious and many people had great hopes for Hilbert's program until Gödel destroyed those hopes with the Incompleteness theorem, which shows that Hilbert's program could not possibly succeed: if the axioms are powerful enough to talk about arithmetic, then they are either inconsistent (you can prove false statements) or incomplete (you can't prove some true statements).

share|improve this answer
2  
If you want to read something about the foundational crisis, I recommend Logicomix: en.wikipedia.org/wiki/Logicomix –  Qiaochu Yuan Jan 4 '11 at 21:53
    
Cool link! –  sova Jan 4 '11 at 22:18
    
In case you haven't seen this, I recommend the following review of Logicomix by Paolo Mancosu from Berkeley. It is useful to separate historical facts from (rather unfortunate) opinions by the authors. –  Andres Caicedo Jan 5 '11 at 0:46
    
Sorry, I forgot to add the link! philosophy.berkeley.edu/file/509/… –  Andres Caicedo Jan 5 '11 at 2:29

Before you understand Gödel's incompleteness theorem, you need to know what a formal system is. A formal system can be thought of as a very strict language where the grammar completely determines the allowable sentences. So, the incompleteness theorem is a statement that "In a formal system sufficiently strong to represent arithmetic, there are sentences which are true but unprovable" which is, to use your phrase, a "limitation on provability (in a formal system)".

You asked "Godel's theorem is proved based on Arithmetic and its four operators: is all mathematics derived from these four operators (×, +, -, ÷) ?"

Answer: No, there is mathematics that does not use those operators. You also are claiming that multiplication and division are inverses on the natural numbers but that is not true. 7/4 is not an allowed division on the natural numbers, it is only when you allow fractions that this is possible.

You asked "There must be operators that are consistent on the natural numbers that we certainly aren't aware of, no?"

Answer: Depends on what you mean by "operator". In mathematics we generally think of an "operator" on natural numbers as a function from a pair of natural numbers to another natural number, so there are in fact infinitely many operators on natural numbers. So yes, there are some we "aren't aware of" because there are more than we could ever count. For example, let's invent the "square product operator" (let's give it the symbol ¤) as taking two numbers to the product of their squares, so 2 ¤ 3 = (2*2)*(3*3) = 36. The possibilities are limitless!

share|improve this answer
    
Thank you very much for your answer. My comment is too long and will come in two parts: I suppose my main conflict is that the library of operators (your definition of a binary operator is what I was thinking when I said operator) limits the provability of relationships. Naturally, relations exist that are not explainable (no axiomatic proof path) by addition and subtraction, and I figured that each operator that can exist is another fundamental piece in trying to define all relations on the set (of natural numbers or reals even). –  sova Jan 4 '11 at 21:52
    
I suppose I understand his theorem to the extent that there are unprovable statements if you have limited axioms -- it seems as though this implies that it is impossible to increase the scope of provability without growing your fundamental operator set. Your "square product operator" is interesting, yet still a subset of multiplication (and addition). I meant operators that are consistent on the natural numbers that are undefinable within existing operators. As you said, there are infinitely many, why not have computers search for new operators/patterns? Do we already? –  sova Jan 4 '11 at 21:59
    
@sova: The important point is that the incompleteness theorem applies to first order systems that include addition and multiplication. Adding operators will not solve the incompleteness. If you follow through the proof, it shows how to make a new unprovable sentence if you add axiom(s) to make the existing one provable. –  Ross Millikan Jan 4 '11 at 22:09
    
@Ross: could you link me to Godel's original proof? –  sova Jan 4 '11 at 22:50
    
@sova: No, but an easy intro is in Nagel & Newman Godel's Proof. It has been in print for 50 years. Much tougher going, but including all the details is Raymond Smullyan's Godel's Incompleteness Theorems. Others will have their favorites –  Ross Millikan Jan 4 '11 at 23:00

This is a personal recommendation.

Peter Smith's "An Introduction to Gödel's Theorems" is very readable and self-contained.

I believe to fully appreciate Gödel's theorems it is a good idea to learn some computability theory before (Smith does include some of this background in his book, but a bit more of perspective doesn't hurt).

Two very good sources for beginners are Cutland's "Computability" and Boolos and Jeffrey "Computability and Logic". There's a forthcoming book from Enderton (who unfortunately recently passed away), "Computability Theory", which I have been able to read and highly recommend too. For a quick (and unusually clear) introduction to computability theory you can take a look at Enderton's notes, which, by the way, are an excerpt from his book:

http://www.math.ucla.edu/~hbe/computability.pdf

I hope this helps (it did to me).

Edit: Let me also recommend a very nice book by Melvin Fitting, Incompleteness in the Land of Sets. As the title suggests, the author shows us how to discuss incompleteness in its most natural place: the land of hereditarily finite sets.

share|improve this answer
2  
I second the recommendation of Smith's book. –  Carl Mummert Jan 5 '11 at 12:15
    
Thank you for your recommendations, they are much appreciated. I went to the library and picked up Cutland's and Boolos & Jeffrey's books, I'm really excited to learn more about this stuff! –  sova Jan 5 '11 at 21:19
    
Good. Be patient, and enjoy! –  Marc Alcobé García Jan 6 '11 at 18:33

Gödel's first incompleteness theorem says, roughly, that a consistent formal system must contain unprovable (but true) statements, even if that formal system only covers integer arithmetic. A stronger system, containing real or complex arithmetic, and sin and cos and so on, must cover integer arithmetic too, so it too must contain unprovable (but true) statements.

share|improve this answer
    
This is very concise. However, it is still not obvious to me that integer arithmetic is a subset of a "stronger system" because it seems as though integers effectively have less detail than the real numbers, and our interpretations of certain operators is that they map to intermediate states (which are only representable by real numbers) that may (in their final state) show integer results. Perhaps this idea makes no sense, but the fact that the integers are a subset of the real numbers does not seem to show that you can't do things on real numbers if you can't do them on ints. –  sova Jan 4 '11 at 22:27
    
To me it's like saying "X is impossible on the spectrum of integers from 1..10 so it is impossible to do X on the infinitely precise spectrum of reals from 1..10" I am really just trying to understand this, and improve my mathematical intuition, so please don't take these concerns as belligerent defense. –  sova Jan 4 '11 at 22:28
    
@sova Your best bet would be to read and understand a summary of the actual proof, then I think you would understand the importance/unimportance of certain details that you have brought up. As Ross said, Nagel & Newman is an excellent start. –  jericson Jan 4 '11 at 23:44

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.