[FOM] natural language and the Fof M
Harvey Friedman
friedman at math.ohio-state.edu
Sat Apr 12 14:17:53 EDT 2003
Reply to Slater 4/11/03 11:49AM.
Slater hints at a theory, following Bunt, with possibly several properties.
1. Set theory is included as a special case. There is a more general
concept of "ensemble", which may or may not be made up of elements.
2. There is a new conditional separation axiom which avoids inconsistency.
3. The theory is some variant of mereology.
*Rather than have FOM readers go find Bunt's book to see if they are
interested, it would be much better if you could briefly sketch the
basics of the theory.*
I am familiar with mereology and David Lewis' idea of interpreting
set theory in mereology. The trouble is, of course, that one has to
augment merology in some significant way in order to do this.
In particular, I think that it follows from the Tarski proof of the
decidability of the first order theory of Boolean algebras given in
Chang/Keisler, Model Theory, that even the first order theory of
successor cannot be interpreted in any consistent extension of the
axioms of Boolean algebra. This almost rules out the development of
any serious amount of mathematics in mereology alone - although one
could conceivably work in some form of mereology which violates one
or more of the axioms of Boolean algebra. It is not clear if this can
be done in any natural or productive way.
In fact, it would be interesting to see just how much of mereology
can be preserved in a theory in the one binary relation symbol
(part/whole) that interprets ZFC.
In order to get around the weakness of mereology from the point of
view of the foundations of mathematics, Lewis introduced a singleton
relation "x is the singleton of y". I have preferred to call it the
naming relation: "x is a name of y", and this has an analogous effect.
A second way to get around the weakness of mereology from the point
of view of the foundations of mathematics is to instead introduce
families of objects (every object is called a "part" my
mereologists), also due to (or at least suggested by) Lewis and
developed by Burgess and others. One can get away with families of
objects rather than relations between objects.
It is well known that one can follow either approach and rather
slavishly write down axioms in this framework that amount to a
translation of the axioms of set theory.
It is not clear what is to be gained by this. One possibility is to
take a different view of this second way. Instead of actually
introducing the families of parts, one can think of this instead as
using monadic predicate calculus with comprehension, so that instead
of objects, one views this as a kind of "logic". This can also be
thought of as "plural quantification" instead of monadic logic. This
is a line taken by Boolos and later Burgess. One can view this as
taking a cue from natural language philosophers to make things more
natural language friendly.
At my visit to Princeton Philosophy Dept last fall, I discovered a
number of surprisingly simple reflection principles that alleviate
any requirement that one slavishly transcribe axioms of set theory
into such contexts, or even in the most bare bones context of all -
just a binary relation epsilon.
My original formulation was in terms of standard mereology with the
"naming relation" - really just Lewis' singleton relation.
This work led to the papers these papers of Burgess:
From Frege to Friedman: A Dream Come True?
http://www.princeton.edu/~jburgess/anecdota.htm
E Pluribus Unum
http://www.princeton.edu/~jburgess/anecdota.htm
Burgess continues to emphasize plural quantification. I haven't yet
taken a view as to how much is gained, philosophically, by this, and
want to look into it further.
I thought it would be simpler to start with just the binary relation
epsilon, and worked hard after returning from Princeton to make
further simplifications. This led to the work reported in my postings
158-161, of 3/31/03. These postings are now in pdf format at
http://www.mathpreprints.com/math/Preprint/show/
One feature of this work in 158-161 is the avoidance of assuming any
form of comprehension. Any form of comprehension that is needed has
to be derived. In the plural quantification approaches comprehension
appears as part of the apparatus.
The earlier work along these lines that I did in the summer of 2002 is in
A Way Out, http://www.mathpreprints.com/math/Preprint/show/
However, this early form of the work isn't nearly as thematic. The
new work is based entirely on various forms of reflection.
I plan to make a posting on the mereological formulation(s). In light
of the new formulations involving epsilon only, in the postings of
3/31/03, the mereological approach might be somewhat simpler than
what I had at Princeton last fall.
In any case, all of this tends to go into the direction of making the
foundations of mathematics more philosophy of language friendly - at
least to some extent.
> Shaughan Lavine, for instance, has produced an axiomatisation
>equivalent to von Neumann's 'which seemed to me to be more natural
>and conceptually based than the usual ones - though I doubt it is
>the last word on the subject' ('Understanding the Infinite', Harvard
>U.P. Cambridge MA 1994, p320, see also pp141-53).
Again, rather than have FOM readers go find Lavine's book to see if
they are interested, it would be much better if you could briefly
sketch it and indicate what is more natural.
>
>Specifically, for instance, y is water not because it is a member of
>a certain set, but because it is part of a certain aggregate (the
>totality of all water), while the formal possibility of a non-set
>equivalent has previously been taken to be a matter of y still being
>a member, but a member of a different type of thing to a set - a
>proper class. The representation of stuff in terms of membership is
>on account of the influence of late nineteenth century thinking
>about number.
The unanswered question is whether anything is to be gained by this
mereological approach to water, for the foundations of mathematics. I
believe that it is philosohically interesting.
More information about the FOM
mailing list