COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	xsre are some suggestions:
C00005 ENDMK
C⊗;
xsre are some suggestions:

1. The paper takes too long before it says what it is all about.
Here is a suggested beginning.

	Artificial intelligence research has experimented with
both procedural reasoning wherein conclusions are drawn by
applying programs to data structures and deductive reasoning
wherein conclusions are drawn by applying rules of inference
to previously accepted sentences.  This paper presents an
rather lengthy example of reasoning in which both processes
are applied.  We have a problem in retrospective chess analysis -
to determine what piece has fallen off a given square of a chess
board from the fact that the position arose in a legal game.
Some conclusions are drawn by routines that look at the given
partial position or hypothetical positions, and others are drawn
by deduction from previous sentences.  The purpose of the example
is to study how these methods of reasoning interact so that
programs using both effectively can be developed.

	The problem being studied is more difficult than can be
solved by any present problem solving program.  In fact, it has
been difficult enough to formulate the known facts about the specific
problem and ghe relevant general facts about chess in first order
logic and in a number of programs called the chess eye.  Our goal
has been to formalize the problem and make a proof acceptable
to FOL, our proof checker for first order logic, that corresponds
to the informal argument that one human gives in persuading another
that his solution is correct.