COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	circum[s84,jmc]		More for circum[f83,jmc]
C00004 ENDMK
C⊗;
circum[s84,jmc]		More for circum[f83,jmc]

Remarks:

	Non-monotonic formalisms in general, and circumscription in
particular, have many as yet unrealized applications to formalizing
common sense knowledge and reasoning.  Since we have to think about
these matters in a new way, what the applications are and how to
realize them isn't immediately obvious.  Here are some suggestions.

	1. When we are searching for the "best" object of some kind,
we often jump to the conclusion that the best we have found so far
is the best.  This process can be represented as circumscribing
⊗better(x,candidate), where ⊗candidate is the best we have found
so far.  If we attempt this circumscription while including
certain information in our axiom ⊗A(better,P), where ⊗P represents
additional predicates being varied, we will succeed in showing
that there is nothing better only if this is consistent with
the information we take into account.  If the attempt to circumscribe
fails, we would like our reasoning program to use the failure
as an aid to finding a better object.  I don't know how hard
this would be.

*****