COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input jmclet
C00005 ENDMK
C⊗;
\input jmclet
\jmclet
\address
Michael Brady, M.A., Ph.D.
Professor of Information Engineering
Department of Engineering Science
University of Oxford
Parks Road
Oxford, OX1 3PJ
England
\body
Dear Mike:

\parindent 20pt

	``Non-monotonic reasoning by the KEW system'' by Anthony Hunter
should not be published in its present form.  Maybe a revised version
could be published in a shortened form.  Here is the basis for this
conclusion.

\item{1.} The paper contains too much exposition of previous work.  The
AI motivation for nonmonotonic reasoning is well known by now.

\item{2.} The author doesn't give much in the way of examples on which the
KEW system has actually been run.

\item{3.} Forbidding using a non-defeasible rule with defeasible premises
is too limiting.  It is reasonable to infer that Socrates is mortal
from ``All men are mortal'' even if the inference that Socrates is
a man is defeasible.  Taken literally, this prohibition prevents
rearranging logical operators.  Thus if we had $p∧q$, proved defeasibly
but a rule required $q∧p$, we would be forbidden to use the logical
rule $p∧q ⊃ q∧p$ as an intermediate step.  I suspect this may not
be the author's intention, but it isn't clear.

\item{4.} There are lots of misprints.  Misspelling Reiter and Gallaire is
only annoying, but there seem to be misprints in important formulas
and places where the terms ``defeasible'' and ``non-defeasible'' are
interchanged.

	A short paper including more results from running the system
might be acceptable.

\closing
Sincerely,
John McCarthy
Professor
\endletter
\end