[FOM] Sharp mathematical distinction between potential and actualinfinity?
V.Sazonov at csc.liv.ac.uk
Mon Sep 29 16:06:41 EDT 2003
"Timothy Y. Chow" wrote:
> Years ago, Stephen Simpson proposed a list of "the most basic mathematical
> concepts," which included number, set, function, etc. I wonder if
> "infinity" belongs on this list?
> It seems to me that at minimum, "potential infinity" of some sort is basic
> to mathematics. Consider the following caricature of a debate that has
> been rehearsed innumerable times on the FOM list and elsewhere:
Although this is a "caricature", it is somewhat misleading.
> P: The concept of N, the set of all natural numbers, is crystal-clear.
> F: No it isn't. For example, first-order PA has nonstandard models.
The point was that there is no such an ABSOLUTE concept as
"the standard model" of PA at all.
> P: Huh? By appealing to mathematical logic, you betray your belief
> that the concept of a *rule* is clear, and N is just as clear as
> the concept of a rule.
> F: Nonsense. A rule is finite, or at most potentially infinite, but
> N is a completed infinity.
The point was that a rule or a formalism or a deduction is not
only finite, but feasible (physically presentable). (It does not
matter here whether N is thought of as completed or only potential
infinity.) Formalisms and the intuitive, imaginary worlds like N
which they "describe" have completely different status if we are
not considering formalisms metamathematically by including them
(by Goedel numbering) into, say PA. We should be quite explicit
about this if we do not want to go into a vicious circle.
Additional essential point is that WE DO NOT NEED any theory of
formal systems to use their rules: we only need to be well trained
for that. (Likewise, we do not need any theory of bicycles to ride.)
These notes allows us to avoid any serious infinite regress in
understanding the nature of mathematical rigorous reasoning.
What is interesting, that the above caricature so strongly
misrepresent the views of F only. Caricature should rather
strengthen some features, may be up to absurd. That would
> I confess that my sympathies lie with P here; I tend to think that
> if we take skepticism towards N seriously, then we also need to take
> "Kripkensteinian" skepticism towards rules seriously, and this leads
> quickly to a wildly, and in my opinion unacceptably, ultraskeptical
> view of virtually all mathematics and logic.
I do not know what is "Kripkensteinian" skepticism. The views
outlined above have nothing ultraskeptical. We just should be
careful enough to not mix everything in one big heap where
we would lost ourselves forever. Is this skepticism or a request
to be sufficiently precise?
Skepticism is rather an antipode of a belief. But should science
be based on beliefs at all? Say, in physics we have some experiments
confirming a law. We can say that this gives us the right to believe
in this law. But more strictly speaking, this means that this is a
good reason to include this law in corresponding physical theory.
Thus, it is not a belief at all. (Will we pray to this law?)
It is rather a hope that this law is reliable. If some couterexample
will be found, there will be no tragedy concerning losted beliefs,
only a problem how to correct the theory. "Belief" is rather
a figure of speach. We can use it, but recall what is really
staying behind this word.
> However, I might change my mind if someone could demonstrate a sharp
> distinction between potential and actual infinity. The distinction
> seems to have pretty much evaporated in modern mathematics, and it
> seems that only the philosophers still talk about it. Or am I just
> underinformed? Is there, for example, a way of drawing a clear
> mathematical distinction between potential and actual infinity that
> blocks the move from skepticism-towards-N to skepticism-towards-rules?
The last line is unclear to me. However, see the above notes.
I do not hope on any possibility of MATHEMATICAL distinction
between potential and actual infinity. Mathematical means to
give a formal definition in a fixed formal system. These ideas
seem too general for that. I hope only on some clarification.
>From my point of view, potential infinity is much less clear
that the actual one. It consists in our (intuitive, imaginary)
ability to always add 1 to any natural number. Also it consists
in our (imaginary) ability to iterate the previous ability.
Say, the first iteration of the successor operation gives rise
to the addition operation x + y. Iterating addition gives
multiplication. Then we can come to the ability to multiply,
to take exponential, any primitive recursion, then Ackermann
function appears, and so on. The main problem is with this
"and so on". We usually slurring over this question. But this
is really the most problematic question on potential infinity
(called also potential feasibility). We could fix this "and so on"
by fixing a (quantifier-free) formalism like PRA (where already
Ackermann function is not presented). But I cannot imagine how to
grasp this "and so on" in its "complete" strength. I even do not
understand what does it mean.
By the way, these considerations already lead us to the idea that
(potential infinite) N is something indefinite. We cannot tell
how "long" it is. Thus, PRA does not guarantee its closure under
the Ackermann function whereas PA guarantees. Everything depends
on the formalism considered. Moreover, all the above informal
considerations on potential infinity are possible ONLY when we
already have some formalisms, each on "an" (not "the") N.
Otherwise we would have no way even to discuss on N in any
non-trivial way (mentioning exponential, etc.).
Taking into account that natural numbers in mathematics are
ALWAYS relativized to a (semi)formal theory, we should conclude
we have no sufficient reason to take the general colloquial term
(natural) "numbers" as mathematically meaningful. It rather denotes
a vague concept in our discussions around mathematics.
On the other hand, actual infinity arises already in Peano Arithmetic
where quantification over "all" natural numbers is allowed together
with the classical logic. Intuitionists and constructivists also
allow quantification over N understood as potentially infinite.
But this is rather subtle point. I am not sure that they do not
come to actual infinity here. If our intuition allows us to
confirm that they "really" use only potential infinity, then
the well-known reduction of clasical PA to intuitionistic or
constructive theories of N would "eliminate" in a sense the
actual infinity from PA.
Actual infinity arises even more explicitly when we consider N as
a "first class citizen" of set theory. This seems clear. I see
potential infinity much more problematic. It arises also in set
theory when we consider, say, the class of "all" ordinals or
cardinals. But I, personally, would rather concentrate on potential
infinity of N.
> FOM mailing list
> FOM at cs.nyu.edu
More information about the FOM