Re: SUO: ELP's summary of MRW's standards experience

["Pierre Grenon" <pierre.grenon@ifomis.uni-leipzig.de>]

John,

Glad you'll work on producing something. Meanwhile, your comments denote some
sort of old school 'monolithism' and few misgivings, imho. It's funny because
what you say on universals and predicates really sounds like it's coming right
out of some fairy-tale told by realists to scare each other. Comments below. 

> Pierre,
> 
> What you're asking for is a very important item on my agenda:
> writing a fairly lengthy analysis and critique of several
> major approaches to ontology, including Cyc, Dolce, and others.
> I wrote a critique of the earlier version of Cyc, which is
> part of the joint review of Lenat & Guha's book (and published
> with several other reviews in AI Journal, vol. 61, 1993).

Yes, it's probably out of date, but that probably have some interesting
hitsotical value at the very least... Thanks for the reference.

> I'll dig it out convert it to HTML and post it on my web site.
> Some of those criticisms still apply, but there have been many
> more changes, most of which are positive, but many of the earlier
> issues still remain.  But there is far too much to itemize in
> an email note.

There's enough traffic on this mailinglist and if you can come up with a
subtsantial review, this is word putting in a file which may be distributed. 

> Just a couple of comments:
> 
> PG >... Looking at a few legacy systems leads you to
> >think for instance that universals may not be spatiotemporal entities.
Somebody
> 
> >made such a remark earlier as if it was an elementary assumption, which is
> not.
> >It is a piece of doctrine.
> 
> No.  That is not doctrine.  It is part of the definition of
> universals -- you either accept it or you choose other words.

Sorry, but no, you are mistaken and that is doctrine. There are many theories
of universals. Not only Plato wrote about universals and under some readings
universals are entities which may be multiply wholly located in space at a
given time and even multiply wholly located in time, some theories allow
intermitent existence in time. References upon request, again I think a
textbook in metaphysics (Lowe as written that kind of texts) will do. (Some
people use an adjective when they want to distinguish between doctrine:
paltonic, aristotelian, conceptual and so on. Look for 'universal in re'.) 

You might be familiar with Armstrong's "Universals and Scientific Realism"
which offers a nice review of several positions available. I think it addresses
some of your terminological problems below. 

> In my case, I prefer to avoid the words "universals" and
> "particulars" because they have accumulated too much associated
> baggage over the years.  Instead, I prefer to use the terms
> "predicate" (or concept and relation type) and "instance".

This is a bad habit. Predicates, concepts, types, and universals are all of
very different kinds of thing. They are related and there are correspondances
but it is lazyness or doctrine (to use a rather neutral term) which leads to
identification. 

Note that predicates are linguistic items while universals or particulars,
assuming you recognize them, are entities. 

> PG> Incidentally, Cyc is in the process of revamping -
> >as cyclists like to say - the treatment of properties. So you'll just get
some
> 
> >confused and contradictory notion that you would have gotten straight from
> a
> >textbook in metaphysics by looking at Cyc or SUMO.
> 
> I use the term "monadic predicate" as a synonym for property.

This is unfortunate. You are bringing too much confusion here and imho you are
mixing language and reality (or model, if the word is offending). 

> And Cyc may be changing some of their properties, but they
> aren't doing anything that changes the equivalence of property
> with monadic predicate.

Please, this makes no sense, precisley because of your confused terminology.
You've just said that you don't distinguish properties and monadic predicates,
why this talk about equivalence then? 

In addition, this is completely opaque to Cyc. You can count all the monadic
predicate in Cyc on a single hand. Cyc uses collections. It is even arguable
whether collections can be understood as 'properties' in the logical sense. 

The revamp which takes place consist in replacing attribute values with
corresponding collections. Neither of them are predicate. Attribute values are
more like Platonic universals. The revamp goes towards some sort of
'#$Collection nominalism'. 

This is not a trivial change in the ontology nor in the language. It might have
no bearing on Cyc's actual inferencing capabilities but this is an entirely
different issue.  

> >People can learn from their own mistake, although not always. It is harder
> to
> >learn from the mistakes of the other. Especially when you don't know what
they
> 
> >are. 
> 
> Right.  That's why I promise to itemize, analyze, and
> criticize their mistakes in my forthcoming report.

I'm eager to read this, but I'm more eager for the next step which hopefully
will go beyond an analysis of the system of shittiness. 

Pierre

> John
> 
-- 
Pierre Grenon
IFOMIS Uni Leipzig
Haertelstr. 16-18
04107 Leipzig
http://people.ifomis.uni-leipzig.de/pierre.grenon/
pgrenon@ifomis.uni-leipzig.de
phone: 49(0)351971672
fax: 49(0)3519716179