clarifying clarifying ontologies

Doug Skuce <doug@csi.uottawa.ca>
Date: Mon, 24 Jul 1995 16:32:10 -0400
From: Doug Skuce <doug@csi.uottawa.ca>
Message-id: <199507242032.QAA09048@mail.csi.UOttawa.CA>
To: cg@cs.umn.edu, srkb@cs.umbc.edu, doug@csi.uottawa.ca
Subject: clarifying clarifying ontologies
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

I normally prefer to listen rather than comment, but I feel the need
to make the following comment on recent debates about ISA hierarchies
and related notions.

I believe the major disagreements that are occurring in this and many
such debates are mainly of three kinds:

1 methodology, i.e. what we are doing and how
2 content (we can't agree on an idea)
3 terminology (we agree on the idea but not the name)

if we ever get this far (in agreement), then we might even be able to agree on

4 axiomatization 

These are often obfuscated, since they interact so much, and result in
much confusion because it is not clear where the disagreement lies. I
feel that there is rarely any disagreement on axioms, because the
actual disagreement is in fact one of the first 3 kinds, or maybe no
one has in fact written any real axioms, if we are using the term
properly. But people THINK it is the axioms they are disagreeing
on. It is probably NOT: the problem really lies earlier, in one of
these other 3 "aspects" of the ontological agreement problem.

So here is the approach I favour, discussed further in my upcoming
paper at the Ontology workshop.

BTW, John Sowa and I have recently discussed this at length and seem
to be in considerable agreement on these ideas.  Our methodologies
seem the same, some of our ideas seem the same, and even (!) some of
our names seem the same. Neither of us has gotten to axioms, despite
years of work, and may not for a while.  The point is we seemed to be
able to separate the first three, largely because there seemed to be
no debate on 1, and we were always clear if we had a 2 or a 3 problem.
(NB this comment is mine, not checked by him. ALso, note the hedge
'seem')

Our methodology then is, roughly: proceed top-down, and try to agree
on fundamental distinctions, i.e to partition our categories according
to some fundamental notions, e.g. 'physical/nonphysical'. The most
critical thing then is to agree on these distinctions, and what
categories they result in, not to write axioms. We therefore strive
first for informal definitions based on these distinctions.  Many
people not only do not need or even want axioms, they would not
understand them. But that should not mean that they can't understand
the concepts and informal definitions.  (Perhaps the term 'axiom' is
in fact being misused, and what some people in these debates may mean
is really what John and I call 'distinctions', a less formal idea,
that underlies axioms.)

He and I are thus are agreed on points 1 and 2, but in this case we
disagree on the term 'nonphysical' (my term) where he prefers the term
'information'. Actually, I think we agreed on a compromise: 'abstract'.

We procede then by trying to identify basic categories in terms of
these distinctions WITHOUT trying to write axioms, only informal
definitions, but done as carefully as we can. We ask questions like:
what is the essence of the distinction, is it universal, does it
create crisp categories or not, is it well known in philosophy, is it
easy to express in major natural languages, and so on. These are by
far the most critical questions, most of which cannot be settled by
axiomatization at all.

Hence formalization in terms of axioms can come only later, when there is
agreement on the first 3 points above. But it is precisely there that
virtually all ontology discussion is mired down at the moment,
partially because of excessive zeal to "axiomize" before "agreeing".

I guess the nub of this is that I don't believe that we need to
axiomatize first as a prerequisite to "agreement". There needs to be a
preliminary step where we clarify the ideas and the terms and the
"design rationale" (a hot idea in design). Of course later
axiomatization may discover certain problems that force rethinking the
ideas and/or terms.

An analogy in the software world would be "writing code" or "formal
specifications" before "clarifying informal requirements". If I am not
clear on how a paragraph is different from a sentence, I should not
start programming a word processor. No good software engineer would
EVER suggest one should do that! (But this is in fact what causes a
lot of the confusion, error, and waste in sofware development.)



Thanks from: Doug Skuce                 | e-mail: doug@csi.uottawa.ca
Department of Computer Science          |
University of Ottawa                    | phone : (613) 562-5800 ext 6686
Ottawa, K1N 6N5, Canada                 | fax   : (613) 562-5187