COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%nonmon.abs[w85,jmc]	Abstract for AAAS
C00004 ENDMK
C⊗;
%nonmon.abs[w85,jmc]	Abstract for AAAS
\input memo.tex[let,jmc]
\noindent {\bf Non-monotonic reasoning}

Logical deduction is monotonic in the set of premises.  Thus if a
sentence  $p$ is deducible from a set  $A$ of sentences and $B$ includes
$A$,  then $p$ is deducible from $B$.  Ordinary human reasoning does
not always have this property, and artificially intelligent systems
also need to supplement deduction by non-monotonic reasoning.
This involves drawing a conclusion from a set of premisses that would
not necessarily be drawn from a larger set.  Since the late 1970s
there have been both expert systems that use non-monotonic reasoning
and mathematical theories of it.  This lecture treats the circumscription
mode of non-monotonic reasoning and some applications to formalizing
common sense knowledge and reasoning.  The basic idea of circumscription
is to assume that certain predicates have minimal extensions compatible
with the premisses.  It is therefore a form of Ockham's razor.
\vfill\eject\end