RE: ontology as science (was: Re: SUO: Re: (ELP's summary of MRW's standards experience))
I support Mike Pool's claim that ontology is not a science. To be "science", it must be falsifiable and it must be possible to devise experiments to test hypotheses indepedently of any particular individual or any particular point in time and achieve the same results. This cannot be done with ontologies because they represent particular interpretations of individual (or communities) of the world that they see - who am I, or anyone else, to tell them that their view of the world is wrong? (Let alone *scientifically* wrong.)
I chimed in a long time ago about the absense of any linguistic, cognitive psychology, or sociology of knowledge content (or ANY human factor content, for that matter) in the SUO development. I think ontologies are as much a "science" as these fields are: they are trying to be "science", but aren't really there yet (IMHO -- my apologies to anyone in those fields on this list.) The lexical aspect of ontologies is mathematics, granted, but the semantic part makes considering it a "science" impossible without considering human factors.
Bill
-----Original Message-----
From: Mike Pool [mailto:mpool@iet.com]
Sent: Wed 6/18/2003 1:03 PM
To: John F. Sowa; standard-upper-ontology@ieee.org
Cc:
Subject: ontology as science (was: Re: SUO: Re: (ELP's summary of MRW's standards experience))
At 02:21 PM 6/18/2003 -0500, John F. Sowa wrote:
>Mike,
>
>In both ontology and nuclear physics, there are many issues
>that are "notational variants". Those are the conventional
>issues that everyone can agree are irrelevant to the substance
>as soon as an appropriate mapping between the two has been
>demonstrated.
>
>> I disagree, or at least I would deny that very many issues
>> in ontology development are serious scientific disputes...
>
>And everything you say in the following discussion would
>apply equally well to ontology and to nuclear physics:
>
>> Nevertheless, this is conventional in a way that issues in
>> nuclear physics would not be (unless, of course, the issue
>> in nuclear physics concerned a stipulation regarding term
>> usage). We may come up with practical reasons for adopting
>> one convention or another but it would be hard to even
>> conceive of devising an experiment that would allow us to
>> determine the fact of the matter as to whether OpenCyc or
>> SUMO "got it right" on the event definition score as it's
>> just sort of odd to propose that one of these definitions
>> is in some sense "falsifiable" (unless we could generate a
>> contradiction from it, but even then it is only falsifiable
>> in combination with the other definitions that allow for the
>> generation of the contradiction). Many ontology issues will
>> concern the ways in which we choose to define terms and,
>> with apologies to Quine and you I guess, this is really
>> nothing like nuclear physics (even if the cleavage between
>> defining terms and figuring out the way the world is isn't
>> very clear).
>
>All the points apply equally well to both:
>
> 1. Both ontology and nuclear physics describe something
> about the world. In that sense, they both require data
> that can either support or falsify the axioms.
Are you saying that ontologists make hypotheses about what exists in the world and then go out and attempt to determine whether they are right? One posits zebras, and then goes out zebra hunting? Another ontologist posits universals and then devises some experiment and checks how many objects are actually instantiations of universals to confirm or nullify her hypothesis? Or on a more concrete level, would I define something called a metre, and then, lo and behold, go out into the world and discover that it is filled with things of one metre in length or with parts that are of one metre in length, would this be how I'd collect data to support my ontological decision? (And what would I say to my American friend who tries to claim that his experiments with the yard produced similar results?)
I really am quite baffled by this claim. Ontologies give us conceptual frameworks with which to talk about what exists, etc. In some sense it's pre or post scientific, but I don't really understand what it would be to collect data on ontology issues.
> 2. Both subjects are developed by people working independently
> who frequently choose different terms and different ways
> of stating the interconnections.
>
> 3. If the terms are simple synonyms (as they often are in
> both subjects), they can be recorded in a simple glossary.
>
> 4. But more often than not, the similarities are not at all
> obvious and deeper analysis is required. One example is
> Schroedinger's wave functions and Heisenberg's matrices.
> It required a great deal of analysis to discover that
> they were equivalent and to demonstrate the mapping
> between them.
>
> 5. And even though they are formally equivalent, physicists
> find it convenient to make use of both representations
> because one version may be easier to apply than another.
>
>All of these considerations have exact parallels in both
>ontology and nuclear physics (or any other empirical science).
>
>Summary: If ontology is a characterization of what exists,
>it must be developed as an empirical science that examines the
>data about what exists.
But how could you ever collect data without ontological assumptions with which to start?
Mike
>It must follow the same scientific
>principles that are used in any other branch of science.
>
>Bottom line: It is foolish to take votes about what exists --
>you have to look, and sometimes you have to look very hard to
>be sure of what you see.
>John