To be honest, this is a question that has been bothering me (provided that I understand this correctly), which probably means I'm not quite sure about what I ask, but I'm looking for some clarification.

I am a graduated mathematician who mostly work on graph theory. I admit I have to learn more about mathematical logic, but when I started to dig deeper, I found (and read) something about statements that are independent of some systems of axioms (not just ZFC: it's just what I heard most, more than PA or others). I read that formally some statements cannot be deduced from a selected system.

I found some independent statements (of ZFC) listed in Wikipedia page, and questions similar to this in Math SE too, but I seem to not get the answer I am wondering about. Some said that when you are trying to prove/deduce a statement, you don't necessarily need to care about its independency or something.

I mean, within my simple, curious mind, which currently might not get the idea of independency at all, isn't it natural to ask:

"Does this mean that if we face a math statement and want to know it is TRUE or FALSE, instead of (usually) only conjecturing that this statement is TRUE or FALSE and providing the reasoning (by this I mean I can try to prove it without any knowledge about axiomatic systems that I use at all), we actually need to also consider the possibility that this statement is independent (of something?) or anything like that?"

I hope anyone gets what I am saying and what my struggles are. This is because when I found some statements that are independent of ZFC or something, I was like "Wait, so does this mean that mathematicians every time must realize that when facing a statement, they must either: a) show it is true, b) show it is false, OR c) show it is axiomatic (can't be proven true or false, relative to a system)?"

I am pretty sure I've never considered option c when doing math (or maybe I have, but I did not care and simply accept that I need AC to show each vector space has a basis). Is that such a normal, and preferable approach? Should we just accept that my above concern will rarely happen? Is mine even a valid concern (since if I read it right, I (generally) should always consider doing additional work when trying to verify a statement)???

Maybe future (and more) researches will lead me to have to consider option c one day, though.

*feel free to criticize this question, if any *some of my tags possibly are wrong ones

  • $\begingroup$ See Examples of unprovable statements : "fter the appearance of Gödel's Theorem a variety of statements more or less directly related to provability were shown to be unprovable in Peano arithmetic and certain other axiom systems. Starting in the 1960s the so-called method of forcing allowed certain kinds of statements in strong axiom systems—like the Continuum Hypothesis in set theory (see page 1155)—to be shown to be unprovable. 1/2 $\endgroup$ – Mauro ALLEGRANZA Nov 30 '18 at 15:23
  • $\begingroup$ Then in 1977 Jeffrey Paris and Leo Harrington showed that a variant of Ramsey's Theorem (see page 1068)—a statement that is much more directly mathematical—is also unprovable in Peano arithmetic." 2/2 $\endgroup$ – Mauro ALLEGRANZA Nov 30 '18 at 15:23
  • 5
    $\begingroup$ When you say “true, false, or independent,” you are mixing metaphors. It is “provable, unprovable, or independent” in a given system. $\endgroup$ – spaceisdarkgreen Nov 30 '18 at 15:45
  • 2
    $\begingroup$ Note that even that third possibility isn't fully satisfactory: there will be statements which are "unprovably unprovable" in ZFC. Basically, there is a sense in which we as mathematicians can never be fully satisfied, or even confident that we will be fully satisfied with respect to a particular problem. Whether that's a positive or a negative depends on one's outlook - personally, I find it quite exciting! But as Carl says, it's a phenomenon one runs into very rarely (and indeed one very rarely needs the full strength of ZFC, or anything close to that), so it's mostly ignorable if one wants. $\endgroup$ – Noah Schweber Nov 30 '18 at 17:54
  • 1
    $\begingroup$ When you are trying to prove something in ZFC, you can always use the fact that any particular statement is true or false. That is, ZFC proves $\theta\lor\neg\theta$ for every sentence $\theta$ (formulated in the language of ZFC). That $\theta$ and $\neg\theta$ might both be unprovable is an entirely different matter. $\endgroup$ – Andreas Blass Nov 30 '18 at 18:52

In a purely formal sense, you are right that if we have a set of axioms and a goal proposition in mind, we could show that the proposition is provable from the axioms, disprovable from the axioms, or independent (neither provable nor disprovable) from the axioms.

However, the phenomenon of independence is much less common when we look at "natural" mathematical statements, and work in strong systems such as ZFC. It is hard to make this formally precise, but if you look a a list of statements independent of ZFC you will see that they tend to have a "set theoretic" flavor. Only the Whitehead problem and the normal Moore space conjecture are sufficiently well known to have an actual name as a conjecture. On the other hand, essentially then entire body of published mathematics is provable in ZFC or small extensions of ZFC.

So, in practice, there's not a good reason for mathematicians working in less set theoretic areas (e.g. finite or countable combinatorics) to worry about whether their conjectures are independent of ZFC. Independence does not happen that often for statements that are encountered "naturally" in practice, but we know that it often happens that a particular result takes a long time and a lot of effort to prove or disprove.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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