COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%ginsbe[f88,jmc]		Errors in intro to Ginsberg's nonmonotonic readings
C00006 ENDMK
C⊗;
%ginsbe[f88,jmc]		Errors in intro to Ginsberg's nonmonotonic readings

Errors in Ginsberg's Introduction - Approaches to nonmonotonic
inference.

p5. The idea that first order logic is adequate.  This is an
oversimplification.  By going to a meta-level, there can be
a monotonic first order theory of nonmonotonic reasoning.  Some
of its sentences might have the meaning, ``If just facts A are taken
into account, then circumscription of  ab  would lead to the
inference that Tweety can fly''.  Very likely this theory, which
no-one has created yet, would not be much less efficient than
the nonmonotonic formalisms themselves - provided the first order
reasoning system had adequate ways of specifying control of
reasoning.  These are not known at present, so Lifschitz's
approach involving algorithms for subcases of circumscription
seems like the right thing to do.

However, the monotonic theory of nonmonotonic inference would not
give the conclusion that Tweety can fly, just something like

circ(ab,A) → flies Tweety.

Suppose that  A  constitutes all that is presently known
relevant to whether Tweety can fly, i.e.  A  constitutes
the database.  Generating from this the assertion,
database(1988.Dec.30) = A is a bit of nonmonotonic
reasoning that no monotonic logic will do.  Once we
have this ordinary monotonic principles might infer
flies(Tweety).

	The suggestion Ginsberg gives lower on the page
that monotonic reasoning can be used if all the qualifications
are stated is inadequate.  A common sense database
cannot list all the qualifications, because they aren't
all known.  This is not just a matter of efficiency.

p8 circumscription is not restricted to minimizing  ab.

p12 - I correctly credited finding the problem parallel to Yale
shooting to Lifschitz in the paper referred to.  See the top of
p. 162 in the Ginsberg compendium.  When I saw the attribution to me,
I worried that I had failed to do so.

p15 - The classification of weak vs. strong was a propaganda
ploy of the expert system promoters.  I dunno about this pessimism.