Benjamin.J.Tilly

What about PA? A collection of simple statements, all of which are manifestly true about the positive integers.

Jan Willem Nienhuys

These statements look simple. But they can do things for you only if you have a vast mathematical apparatus that includes:
  1. conventions regarding names of variables;
  2. conventions regarding scopes of thoses names;
  3. rules about what kind of substitutions are permittable;
  4. conventions regarding definitions;
  5. conventions regarding what is considered: sets, functions, products.
  6. rules of inference (i.e. logic).
I just mentioned in my previous post 1 (variable names). If one proposes to name variables x, x', x'' and so on, one certainly needs to know that all these names are different. Of course, in my heart I know they are different. But if the formal system is to be absolutely watertight, I must prove that they are different. In other words, I must apply the Peano axioms to the language used to talk about them.

Benjamin.J.Tilly

I think that you are going overboard. For instance at first all that you need to use are PA with a finite number of variables. The other conventions are taken care of by first-order logic. (A series of rules that are again very reasonable. If I was to learn that first-order logic was hopelessly flawed then I would be severely disturbed.)

If you are really concerned then you can use a "boot-strap" technique so that statements which involve an arbitrary number of variables are talked about in a system that uses only finitely many of them. Then you can talk about several groups of variables, each group of which could have any number of variables in it, and so on ad nauseum. The principle would be like the distinction between first-order logic, second-order logic, ect. In principle this could get rid of the application of the same version of Peano's axioms to the version that you are using at any point.

Personally I do not see that there is any point in bothering. Foundations, after all, at some point come down to belief. So what do you believe in? Personally I accept PA completely, and ZF seems reasonable. Beyond that I do not worry about it since I do not personally use it...

Jan Willem Nienhuys

The fact that my experience with nonnegative integers makes me believe that the Peano axioms are manifestly true is irrelevant. My experience certainly doesn't include having often verified that adding 1 consecutively 10^80 times gives each time a different result.

The basic trouble with the Peano axioms is that they implicitly assert that any set of positive integers has a smallest element. If you think of simple Ramsey numbers, then you'll realise that even in simple cases this is more a metaphysical assertion than a matter of "manifestly true". How can I have confidence that any set of positive integers, no matter how unwieldy its definition has a smallest element? The unwieldyness of definitions is made possible by having an unlimited amount of positive integers (and variable names) at your disposal.

Benjamin.J.Tilly

Well, if you want to get nasty about it, there is an infinite set of positive integers such that no positive integer can be proven to be in it in your (consistent) formal system. (The set's definition depends on what formal system you are interested in, but that can be rectified in that I can define it in such a way that, for instance, no formal system stateable in under a hundred pages of writing can verify that any number is in the set.)

If things like this bother you, then become a constructivist.

And as for the Ramsey numbers? I have no problem with those. I know that there is an answer. I know how to find that answer. That this is impractical is a matter of practice, not principle. (Then again, I am a mathematician. :-)

Jan Willem Nienhuys

The Peano axioms plus a formal system in which they are embedded must be free of contradictions, i.e. for no statement A both A and not-A may be provable. (Note that I apparently want to allow variables like A to denote statements; the formal system should have watertight descriptions - no handwaving or illustration by example allowed - of how to go about making variables represent statements.)

Only such a system as a whole can conceivably contain a contradiction. Without the formal framework a contradiction doesn't make much sense.

Benjamin.J.Tilly

Ummm...I think not. There are (as has become apparent in some discussions) many ideas that make a great deal of sense, whether or not they have been formalized. For instance things like Russell's paradox, the exam paradox, and the barber paradox all make sense to people and can generate discussion even without a formalization.

Personally I think that mathematicians are sometimes too hung up on the details of formalizations and lose track of the actual math they are talking about. As the old saying goes, mathematicians are platonists on weekdays and formalists on Sunday. I personally think that this tendancy is a good and healthy one!


Back to Part 6
Return to Menu
Go On to Part 8: Sublime Confidence

Сайт управляется системой uCoz