[ont.events] Temporal Reasoning: What's the Problem Anyway?

ylfink@water.waterloo.edu (ylfink) (12/13/88)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

ARTIFICIAL INTELLIGENCE SEMINAR

                    - Wednesday, December 14, 1988

Mr.  Scott D. Goodwin, Department of Computing Science,
University   of   Alberta,  will  speak  on  ``Temporal
Reasoning: What's the Problem Anyway?''.

TIME:                1:30 PM

ROOM:              DC 1304

ABSTRACT

Much   interest   has   been  focused  on  nonmonotonic
reasoning in temporal domains since Hanks and McDermott
discovered that intuitive temporal representations give
rise  to the multiple extension problem.  In this talk,
nonmonotonic   reasoning   in   temporal   domains   is
considered   from   the  perspective  of  the  Theorist
hypothetical   reasoning   framework.    Theorist   was
originally  developed at the University of Waterloo and
now  several  researchers  across  Canada  are actively
investigating its foundations and applicability.

The  first  part of the talk will describe the Theorist
framework.   We  then  show how Theorist can be used to
represent  temporal  domains  (in  the case of discrete
time).   We  discuss the multiple extension problem and
how  it  arises  in  Theorist.  One attempt to overcome
this  problem,  a  theory  preference scheme called the
chronological  maximization  of persistence (CMP), will
be   examined.    We  show  why  this  scheme  is  only
applicable under certain circumstances.

The  final  part  of  the talk is more speculative.  It
describes  a  more  robust  approach  to  the  multiple
extension  problem currently under investigation.  This
approach  adopts  the  view that in many cases multiple
extensions  arise  because  of  failure  to  take  into
account   what   the   probabilists  call  conditioning
information.   Finally  we  describe  how  some  of the
problems  for  which  CMP  is  not  applicable  can  be
addressed  using a theory preference criterion based on
Markov's principle.