COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	reify[e82,jmc]		Reification and natural kinds in AI
C00005 ENDMK
C⊗;
reify[e82,jmc]		Reification and natural kinds in AI

	This paper explores the idea that the first order logic
formalisms (usually referred to as "predicate calculus" in the
AI literature) used so far in AI don't  reify  enough and mistakenly
attempt to define many concepts that should be taken as
natural kinds.  Our criticisms seem also apply to philosophers, but
we won't discuss here issues in the form they have been raised in
the philosophical literature.

	We will explain our use of  reify  and  natural kind,  so
it won't be necessary to consult the philsophical literature just
in order to read the paper.

	Many of the ideas arose through introspection and involve
conjecture about how people, especially children, for and use 
certain classes of concept.  Therefore, criticisms will probably
occur to people with different introspections or who have actually
studied how children form and use concepts.  Since our object is AI,
not child psychology, the most pertinent criticisms will be those
asserting that programs based on the ideas of the paper won't
behave as hoped, e.g. won't achieve their goals in a wide class
of situations.  Moreover, since our present goal is a modest amount of
intelligence, it would be most relevant to find something simple
rather than something exotic that the programs won't be able to do.
(Since this is a theoretical paper, there isn't a program that can
be criticized directly).  In fact, the critical reader should ask
himself what is the very simplest thing that the programs won't
be able to do.