Axiomatic set theory builds up set theory from a set of fundamental initial rules. The most common axiomatization, which we’ll be used, is the ZFC system: Zermelo-Fraenkel with choice set theory. The ZFC axiomatization consists of 8 basic rules which are pretty much universally accepted, and two rules that are somewhat controversial – most particularly the last rule, called the axiom of choice.
Naive set theory is fun, and as we saw with Cantor’s diagonalization, it can produce some incredibly beautiful results. But as we’ve seen before, in the simple world of naive set theory, it’s easy to run into trouble, in the form of Russell’s paradox and a variety of related problems.
For the sake of completeness, I’ll remind you that Russell’s paradox concerns the set R={ s | s ∉ s}. Is R∈R? If R∈R, then by the definition of R∉R. But by definition, if R∉R, then R∈R. So R is clearly not a well-defined set. But there’s nothing about the form of its definition which is prohibited by naive set theory!
Mathematicians, being the annoying buggers that they are, weren’t willing to just give up on set theory over Russell’s paradox. It’s too beautiful, too useful an abstraction, to just give up on it over the self-reference problems. So they went searching for a way of building up set theory axiomatically in a way that would avoid problems by making it impossible to even formulate the problematic statements.
So, what’s set theory really about?
We’ll start off, for intuition’s sake, by talking a little bit about what’s now called naive set theory, before moving into the formality of axiomatic set theory. Most of this post might be a bit boring for a lot of you, but it’s worth
being a bit on the pedantic side to make sure that we’re starting from a clear basis.
A set is a collection of things. What it means to be a member of a set S is
that there’s some predicate PS – that is, some way of describing things via logic – which is true only for members of S. To be a tad more formal, that means that for any possible object x, PS(x) is true if and only if
x is a member of S. In general, I’ll just write S for both the set and the predicate that defines the set, unless there’s some reason that that would be confusing and/or ambiguous. (Another way of saying that is that a set S is a collection of things that all share
some property, which is the defining property of the set. When you work through
the formality of what a property means, that’s just another way of saying that there’s a
predicate.)
While I’ve been writing about the Surreal numbers lately, it reminded me of some of the fun of Set theory. As a result, I’ve been going back to look at some old books. Since I’ve been enjoying it, I thought you folks would as well.
Set theory, along with its cousin, first order predicate logic, is pretty much the
foundation of nearly all modern math. You can construct math from a lot of
different foundations, but axiomatic set theory is currently pretty much the dominant approach. (Although Topoi seem to be making some headway…)
There’s a reason for that. Set theory starts with some of the simplest ideas, and extends in a reasonably straightforward way to encompass the most astonishingly complicated one. It’s truly remarkable in that – none of the competitors to set theory
as a foundation can approach the intuitive simplicity of set theory.
So I’m going to write a bit about set theory as I explore my old books. And I thought that a good place to start was Cantor’s diagonalization. Cantor is the inventor of set theory, and the diagonalization is an example of one of the first major results that Cantor published. It’s also a good excuse for talking a little bit about where set theory came from, which is not what most people expect. Set theory was originally created as a tool for talking about the relative sizes of different infinities.