COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%frame.abs[w87,jmc]	Abstract for Kansas conference
C00005 00003	\centerline{This draft of \jobname .abs[w87,jmc]\ TEXed on \jmcdate\ at \theTime}
C00006 00004
C00007 ENDMK
C⊗;
%frame.abs[w87,jmc]	Abstract for Kansas conference
\input memo.tex[let,jmc]
\title{The Frame Problem Today}
\noindent Abstract: The frame problem is that of specifying what doesn't
change when an event occurs.  It is readily solved in systems that have
fixed sets of actions and fixed sets of fluents built into the program.
It is most acute when the system must be open-ended, i.e. must be ready to
accept descriptions of new kinds of events and new kinds of fluents whose
values are in general not affected by events whose descriptions don't
mention them.

	Our formalisms are in languages of mathematical logic, mostly
first order.  We begin with a 1960s situation calculus description of the
effects of moving and painting objects with explicit frame axioms.  We go
on to discuss frames as objects and introduce a formalism that describes
change as making generalized assignment statements using my 1963
axiomatization of assignment.  We discuss the length of proof that a
fluent hasn't changed its value after a large number of events have
occurred and discuss sufficient conditions under which this takes a
constant number of proof steps independent of the number of events that
are presumed to have occurred.

	Next we discuss non-monotonic formalizations of the effects
of actions and their use in solving the frame and qualification
problems.  The relation between model-theoretic treatments and
second order axiomatizations is discussed.  There is some discussion
of the relation between the frame problem and the heuristics of planning.
\centerline{This draft of \jobname .abs[w87,jmc]\ TEXed on \jmcdate\ at \theTime}
\centerline{Copyright \copyright\ \number\year\ by John McCarthy}
\vfill\eject\end