Re: Multiple Inheritance and "rigid" properties
At 9:49 AM +0100 6/21/00, martin_king@UK.IBM.COM wrote:
>Dear all. a contribution to this debate spurred by Nicola's.
>
>1. I would like to offer the strongest possible caution against believing
>that there are in any absolute sense classifications which are somehow more
>rigid than all others. Note that this does not mean I am against trying to
>define and seek agreement on some classifications to be treated as more
>rigid than others, but that having done so we cannot claim they are more
>rigid except in so far as some body have agreed to treat them so.
Two things in here. First, I don't think you get the notion we are
trying to capture with the word "rigid". A rigid "classification"
(your term) is one that is essential for all it's instances. We have
not yet begun to really discuss what the concepts in our SUO are, but
defining which ones we take (we the SUO designers) to be rigid will
be important, I believe. Take this simple example:
At least in the normal account, the property PERSON is rigid. At the
moment I'm not interested in whether you agree that PERSON is rigid,
at the moment I'm interested in WHAT IT MEANS for a property, e.g.
PERSON, to be rigid. I say it means that no instance of PERSON can
cease being an instance of PERSON, unless it (the instance) ceases to
be. Being a PERSON is essential to the existence of all instances of
person. There is a modal necessary there, for those that know what
that means.
Now, in the normal account, the property STUDENT is not rigid. This
means that it is possible for an instance of STUDENT to stop being an
instance of student.
Again, at the moment I'm not claiming PERSON *is* rigid and STUDENT
is not, I'm claiming that IF WE CHOOSE to say PERSON is rigid and
STUDENT is not, then the above holds. OK?
What we have, then, is a way of analyzing properties and describing
part of what we intend them to mean for our purposes.
Second, of course I agree that what we will come up with is
subjective. However, you're argument is actually one about trying to
devise an SUO at all. No matter what we come up with, there will be
people who don't like it, there will be cases that won't fit. But I
take for granted in this list that we will try to come up with an SUO
- and given that effort, we should try to make every aspect of our
intended meaning as clear as possible. Rigidity is one of these
aspects.
> One
>criterion for such a set of "rigid" classifications might be that they
>minimise the need for multiple inheritance. I also believe it is
>inherently difficult to agree on such classifications. For justification
>of these beliefs, I strongly recommend "Women, Fire, and Dangerous Things"
>George Lakoff, 1987, or to show that this debate is even older,
>"Understanding Computers and Cognition", Terry Winograd and Fernando
>Flores, and "Data and Reality", William Kent.
The debate is far older than that. Pick up a few books in Philosophy
and start going backwards. But again, that there is a debate, and
an OLD one, is an argument against having an SUO at all. To me
that's like saying, (in the US) we have democrats and republicans and
they rarely agree on anything, so why vote?
-Chris
****NOTE NEW AREA CODE:
Christopher A. Welty http://www.cs.vassar.edu/faculty/welty/
Vassar College Computer Science Dept. Voice: (845) 437-5992
Poughkeepsie, NY 12604-0462 Fax: (845) 437-7498