On Mon, June 3, 2013 13:30, Barkmeyer, Edward J wrote: (01)
> I believe that for making an ontology for a particular purpose, one can
> encode a great many 'facts', that are "taken to be true" for the purposes
> of reasoning with that ontology for that purpose. ... In particular,
> the knowledge engineer must make a separate judgment as to such an
> ontology is appropriate for some other purpose. (02)
Which is why it is useful to have a basic sharable ontology that does not
go into detail, and various theory ontologies with rules (or other
assertions)
describing how things work in certain classes of context. (03)
> Some notes below.
>
> John F Sowa wrote, on June 02, 2013, at 8:12 AM: (04)
>> In various notes to Ontolog Forum, I emphasized the importance of laws
>> (both physical and social) as a foundation for ontology. I won't go
>> into all the
>> details, but I'd like to make a few comments:
>>
>> 1. Laws of physics provide a more fundamental explanation of many
>> properties than the notions of "disposition" or "tendency".
>> Physicists, for example, would explain why glass is fragile
>> along the following lines:
>>
>> "Look at the structure of the material and how the atoms and
>> molecules are linked to one another. Given that structure
>> and the laws of quantum mechanics, you can predict that
>> under certain conditions,
^^^^^^^^^^^^^^^^^^^^
This is the key phrase. Certain rules are valid in certain contexts. (05)
>> something made of steel will bend, and
>> something of the same shape but made of glass will break." (06)
> ...
> So, while this idea is appealing, physicists would be careful to say that
> they can accurately predict the stress resistance of certain glass
> structures and certain steels, but possibly not others. (07)
Yes. Predicting these from first principles is often nigh impossible. (08)
> There are several
> steels that will shatter at temperatures near 0 degrees Kelvin, for
> example. (09)
I'm sure that such temperatures are outside the "certain conditions"
that John was referring to. (010)
> ... So, your example "laws" could probably be used
> to draw appropriate conclusions for common
> building materials under typical usage conditions,
> but would be inappropriate for [other] conditions. This is the
> danger. (011)
Ed, i guess that you see the danger as being people using rules
outside of their scope of applicability (i.e., the context in which
they are valid). This suggests that the context for any set of
rules should be specified, and that anyone intending to use such
rules should verify that they are within a context in which the
rules are defined to be valid. (012)
>> 2. [the term 'disposition'] ... (013)
>> 3. For the social sciences, the interactions are far more complex
>> than in physics. But there are regularities that can be used to
>> make predictions that have a high probability of being correct. (014)
> Aye, and there's the rub. You are now talking about Bayesian inferences,
> which is significantly different from "common logic" and what most of us
> think of as "ontologies". (015)
An ontology can model any logic, including Bayesian. (016)
> Most of these social examples relate to
> "business rules", rather than "axioms". A "business rule" can be
> "violated" (fail to hold), without creating an inconsistency in the
> knowledge base. ... (017)
> As I said, I don't disagree that you can make an ontology that assumes
> these rules as axioms, but you have to know that the assumption is not
> detrimental to the purpose of the ontology. (018)
Agreed. (019)
If rules are defined to be default true, instead of absolutely true, one
can often come up with arguments for and against a certain conclusion. (020)
There are many techniques (including fuzzy reasoning) for coming up
with answers, but one must always allow for, one moves from
a probabilistic answer to what happens in a specific case, the possibility
of the probable answer not corresponding with reality. (021)
>>...
>> In summary, I believe that laws of nature and social behavior are a
>> better foundation for ontology than dispositions.
>> I would also recommend game theory as a useful methodology
>> for reasoning about social behavior.
>
> I agree with the first sentence, but with all the caveats above. We have
> theories that are based on observations and that are accepted because they
> produce predictions that have been verified by experiment. We can
> therefore take these theoretical characterizations to be 'true' laws of
> behavior, for the purpose of reasoning about a particular class of
> situations, as long as that assumption is not seriously harmful to our
> purposes when it proves to be an oversimplification.
>
> ... A game theoretic model of a social interaction
> and an axiomatic model of a social interaction are
> radically different models. (The axiomatic model -- the
> "ontology" -- is the primary subject of this forum.) (022)
I would not equate the axiomatic model with the ontology. An
ontology can model a game theoretic model. (023)
> If the game theoretic model does not have a 'saddle point',
> the optimal game theoretic solution is a mixed strategy:
> When X happens, do Y 60% of the time, and
> do Z 40% of the time (randomly). (That is how hedge
> funds work.) Axiomatically, that reads, if Xthen What?. (024)
If X
then
if (60% of the time)
Do Xaction1
else
Do Xaction2
else <do nothing>. (025)
Use your own function to generate 60% of the time. Neither
the ontology nor the KB need to see inside that black box. (026)
-- doug foxvog (027)
> ...
> -Ed
>
>> John (028)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (029)
|