[Top] [All Lists]

Re: [ontac-forum] John Sowa

To: ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Thu, 18 May 2006 08:27:35 -0400
Message-id: <446C6837.6080505@xxxxxxxxxxx>
Barry and Christopher,    (01)

Yes, I admit that I was using a rhetorical technique
to emphasize the point and that the word "metaphysics"
is a later coinage:    (02)

BS> Aristotle never used the term 'metaphysics' (which was
 > almost certainly introduced by the librarians of Alexandria,
 > for those treatises which came after the treatises in
 > physics on their library shelves). Rather, Aristotle used
 > the term 'first philosophy', which rather suggests that he
 > was on Azamat's side.    (03)

Nevertheless, the so-called "first philosophy" never comes
first in a child's early stages of language learning.  When
children learn new words, it is always from a posteriori
experience, not from a priori assumptions.  Even in mathematics,
in which the axioms are assumed a priori before the deduction
begins, the discovery methods used by mathematicians to find
new axioms are based on analogies with a posteriori experience.    (04)

BS> How, I wonder, would John have responded to Newton, had he
 > been around when he published his Principia Mathematics?
 > "Oh, Isaac, we don't disagree about the goals.  My only point
 > is that we still have a long way to go before we reach them.
 > Perhaps  we might get there in a few more centuries, or perhaps
 > it may take billions of years.  Nobody knows. ...'    (05)

My attitude toward any new theory that explains a wide range
of data is the same:  It's a great achievement.  But as we have
seen, no great theory lasts forever.  And with the quickening
pace of modern science, a long lifetime for a theory is closer
to 30 years than 300 years.  An example is the "standard" model
for the subatomic particles, which was formulated in the '70s
and which is now seriously threatened by the discovery that
neutrinos have mass.  As I said, nobody knows whether any
empirical hypothesis will stand the test of time.    (06)

CS> Isn't there a middle way?  How about an upper ontology
 > for human-universal pragmatics (if I may risk miscommunication
 > with that choice of words...)?  Something like this :-
 > The upper ontology asymptote we might first seek would be
 > one which frames the overall structures, uniformities,
 > patterns, laws and constraints in the human knowledge discovery
 > and creation process as we currently find it, so as to better
 > enable collaboration (better than the previous version, that is)
 > between all the various overlapping and intermingled bodies of
 > people presently on planet Earth.    (07)

That's a very important point.  When I finished my KR book in
1999, I still had hopes for a conventional ontology with a fairly
stable upper level.  But all the wrangling in the SUO community,
with competing groups demonstrating that they could pick and choose
different upper levels while still developing useful lower-level
theories, convinced me that Doug Lenat was right:  the upper level
is much less important than the middle levels, and most of the
axioms necessary for any practical problem are in the lowest
levels of the microtheories.    (08)

I was also impressed by the fact that the wrangling groups could
all agree on aligning their so-called "formal" ontologies with
the very informal WordNet, which was derived from the vague and
ambiguous natural language that Frege, Russell, and Carnap were
trying to replace.    (09)

I have always had a much higher respect for NLs than the Fregean
crowd, but my respect has been steadily increasing.  In fact,
I would now claim that the vagueness and ambiguities of NLs are
certainly not defects, but essential properties of any medium of
expression that can, as Christopher says, "enable collaboration...
between all the various overlapping and intermingled bodies of
people presently on planet Earth."    (010)

Therefore, my current recommendation (which I have been stating
in various ways for the past year or two) is to have a sparsely
axiomatized type hierarchy that would be closer to a cleaned-up
version of WordNet than to any of the popular proposals for
formal ontologies.  In fact, I would say that the upper level
should have no empirical axioms whatever.  The only axioms
should be what Carnap called "analytic" -- the definitions or
"meaning postulates" that distinguish one type from another.    (011)

In this sense, the upper levels could be called "first philosophy"
because they could never be falsified by any a posteriori empirical
facts.  The lower levels, which may contain many axioms derived
from empirical data, would draw upon the upper levels as a source
of useful distinctions.  But to allow for the inevitable revisions
that take place with progress in science and engineering, the
lower levels would be compartmentalized in a wide range of problem
or task-oriented microtheories.    (012)

As for collaboration and interoperability, both people and computer
systems never require a global alignment of all their categories.
It is only necessary to agree on those low-level categories that
characterize the specific problem or task.  That is how people
have been interoperating for millennia, and that is how computer
systems have been interoperating for the past 50 years.    (013)

In short, to promote interoperability, focus on the messages that
two agents (human or computer) use to communicate about the task.
Then align the microtheories that characterize those messages.
Categories that are irrelevant to the task need not (and probably
should not) be involved in the process of alignment for that task.    (014)

John Sowa    (015)

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (016)
<Prev in Thread] Current Thread [Next in Thread>