Re: clarifying clarifying ontologies

fritz@rodin.wustl.edu (Fritz Lehmann)
Date: Wed, 26 Jul 95 20:09:58 CDT
From: fritz@rodin.wustl.edu (Fritz Lehmann)
Message-id: <9507270109.AA06870@rodin.wustl.edu>
To: cg@cs.umn.edu, doug@csi.uottawa.ca, phayes@cs.uiuc.edu, srkb@cs.umbc.edu
Subject: Re: clarifying clarifying ontologies
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

     Pat Hayes wrote:
---begin quote---
Well Doug, I guess I couldnt disagree more. I think your position is based
on a fundamental error, which is that the way to get formalizations is to
talk and talk until our thoughts are Clear, and then just jot them down in
a handy logic. The reason this doesnt work is one you may not have
discovered if you have never gotten as far as the last stage of
axiom-writing: the process of writing the axioms (or frames or whatever)
reveals distinctions, errors and inconsistencies which you simply hadnt
thought of before. It is impossible to get our intuitions clear just by
informal philosophizing, no matter how carefully done.
     At the risk of sounding repetitive, I can cite temporal relationships,
...
---end quote---

     Well, mushy philosophizing at the upper end is one hazard to
be avoided, but getting mired in formal details at the lower end is
another hazard to be avoided.  For example, Hayes himself has pointed out that
almost all of those formal incompatibilities among his time axiomatizations
are IRRELEVANT to almost all practical, human-scale activities -- they
arise only in the extreme borderline cases, where EXACT identity of
time periods is involved, for example.  They involve issues of
a discrete time-line of tiny moment-points versus a dense line
like the rationals versus a supposedly Cantor-continous line like
the Reals.  Joe Blow, who needs ontologies of business systems,
calendars, etc. couldn't care less, particularly since all time
measurements, like other physical meaurements, are necessarily
approximate anyway.  It is only in the rarefied world of precise
logic and mathematics that these "continuum" problems even arise
at all.  That's a _drawback_ of typical logical axiomatization.

     I'm in favor of both kinds of ontologizing, top-down and
bottom up.  I'm even more in favor of figuring out the middle
levels, which is where I suspect the real knowledge bottleneck
is, and grounding the upper levels in the lower levels.  Hayes
and Skuce are attacking different, legitimate ends of a big 
problem.

     (Meanwhile, we really do need a servicable theory of
approximation.  Tolerance theory, Topaloglou's "haze space/time",
Cohn's and my EGG/YOLK theory, the method of the Rome Planning
Ontology, and ideas from Measurement Theory may help.)

                          Yours truly,   Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-856-0671               email: fritz@rodin.wustl.edu
=============================================================