[FOM] ZC vs. ZFC
Thomas Forster
T.Forster at dpmms.cam.ac.uk
Sun Jan 31 07:43:51 EST 2010
We seem to be having this discussion about the need for replacement
over and above Zermelo far too often for my taste. Here is my
two penn'orth; i shall be brief, and then silent!
The point is often made that replacement doesn't seem to be needed
for much of `ordinary mathematics'. This isn't really a point greatly
worth worrying about even if it is true, since what is ordinary
mathematics in 1900 isn't ordinary mathematics in 2000 or 3000 and
whereas it might be worth worrying whether or not replacement is *true*
(whatever that might mean) in the cumulative hierarchy, worrying about
where it is needed in this place at this time is not a mathematical
concern: mathematics is time-invariant. (To be fair, some of the people
who observe from time to time that replacement doesn't seem to be needed
&a, &a, do also say that is not a point of metaphysical significance).
Harvey Friedman has a catalogue of nice facts about sets of low rank that
can be proved only by reasoning about sets of high rank. This *ought*
to shut down the debate, but for some reason it doesn't. I think it is
something to do with the distain in which logic is held by many
mathematicians, which makes them reluctant to accept illustrations of the
kind Harvey supplies. Another reason for their failure to appreciate the
need for replacement is a general disinclination to pay attention to what
axioms are used, so that even when replacement *is* needed for what they
do, they don't realise they are using it. (This point is commonly made
about AC, but it holds for replacement as well)
This brings me to what - for me at least - is one of the major reasons for
accepting replacement. There are at least three halfway-sensible ways of
thinking of natural numbers as sets that have been proposed: there are the
Zermelo naturals, the von Neumann naturals, and there are Scott's trick
naturals. It is a fact so important and so basic that it tends not to get
spelled out that *it doesn't matter which implementation we use*.
Certainly the von Neumann implementation is cuter than the others, and
more generally used, but we must not lose sight of the fundamental insight
that *it shouldn't matter which one we use*. Actually i don't think we
are in danger of denying this point, so much as *overlooking* it. And
here is the key point:
*** If we want to maintain a stance of lofty indifference about ***
*** our choice of implementation then we have to assume replacement. ***
And lofty indifference is surely the mathematically correct stance.
Perform the following thought-experiment: Ramanujan turns out not to have
been dead, but in a coma, and he wakes up. (You will remember that Hardy
said of him that all the natural numbers were his personal friends.)
When he opens his eyes, you greet him with the news that during his
absence we discovered that the natural numbers were hereditarily
transitive finite sets wellordered by the membership relation. Will he
say
(i) ``Thank you for sharing that with me''
or
(ii) ``Security, please take this lunatic away''
I think we can all agree that the fact that natural numbers were
hereditarily transitive finite sets wellordered by the membership relation
is of no conceivable use to a number theorist.
If you propose to refrain from using replacement then you cannot have
lofty indifference. Notice that this means that you have to plump for one
(or at least a restricted bundle of) implementations of (e.g., ordinals)
and decide that ordinals just are sets of that kind! How ironical that
the people who regard replacement as an unmathematical logicians' plaything
are denying themselves the right to the lofty indifference that many of
them (I have in mind for example Godement's *Cours d'algebre* - thanks to
Adrian Mathias for drawing it to my attention) fairly explicitly come out
with:
If you want to be able to behave like an - ``ordinary!'' - mathematician
you have to have replacement.
URL: www.dpmms.cam.ac.uk/~tf; DPMMS ph: +44-1223-337981;
UEA ph: +44-1603-593588 mobile in UK +44-7887-701-562;
(Currently in the UK but mobile in NZ +64-210580093.
Canterbury office fone: +64-3-3642987 x 8152)
More information about the FOM
mailing list