COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	The problem of making computers carry out common sense reasoning was
C00005 ENDMK
C⊗;
The problem of making computers carry out common sense reasoning was
first posed in [McCarthy 1959], and some results were reported in that
paper and in [McCarthy 1963] and [McCarthy and Hayes 1970].
Because the problem has proved very difficult and is vital for
artificial intelligence, it is important to try to split it
into subproblems that can be attacked separately.  We have already
mentioned the split into heuristic and epistemological parts.  Now
it turns out that the epistemological part can be further split into
the subproblems of discovering the facts of the common sense world
and the problem of following, i.e. checking, common sense reasoning.

Our last publication on the common sense facts is [McCarthy and Hayes 1970],
and since that time additional results have been obtained on how
to express knowledge and the problem of simultaneous action, but it is
not quite time to redo the whole formalism of that paper.  

We can isolate the problem of following common sense reasoning from the
problem of finding the basic common sense facts by finding a domain in
which the facts are clear, but the reasoning is closer to real world
common sense reasoning than to mathematical reasoning.  Such a domain
is provided by certain chess problems in which the tree search
traditional in chess programming has to be combined with the step-by-step
deductive reasoning characteristic of mathematical logic, and both
have to be supplemented by a mode that corresponds to observing the
chessboard.  This mode