COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub" source
C00007 ENDMK
C⊗;
.require "let.pub" source;
∂AIL Dr. Stevan Harnad↓%2The Behavioral and Brain Sciences↓P.O. Box 777
↓Princeton, New Jersey 08540∞

Dear Dr. Harnad:

	Thanks for your letter of June 17.  I have two possibilities
for a %2BBS%1 target article.

	The first is based on a paper entitled
%2Ascribing Mental Qualities to Machines%1 which appeared in Martin
Ringle's %2Philosophical Perspectives in Artificial Intelligence%1,
Humanities Press.  The publisher seems to have done nothing with the
book, and I don't even have a copy.  I enclose a slightly
updated version, but I would want to update it much more to take into
account developments in non-monotonic reasoning, to take into account
philosophical work I have come across recently (even though some of it
is old), to correct obscurities that have been pointed out, and to
give better examples.  It also appeared as a report of our laboratory,
but that version was based on the wrong computer file and contains
errors.

	I believe the paper is suitable for %2BBS%1 treatment, because
it shows that successful AI requires going beyond what most philosophers
(except for Daniel Dennett and followers), many AI people,
and most behavioral scientists
will allow in ascribing mental qualities to machines.

	The second possibility is a review article on non-monotonic
reasoning.  My recent article on that subject presents
just one approach, and I think the %2BBS%1 audience would be better
served by a comprehensive article.

	Non-monotonic reasoning has the property that a conclusion %2p%1
non-monotonically obtained from a collection %2A%1 of premisses may
not follow from a collection %2B%1 that contains all the sentences of
%2A%1 and more.  Logical deduction is monotonic, but humans do and
intelligent machines must reason non-monotonically as well.  I enclose
a copy of a report based on a paper published in %2Artificial Intelligence%1,
that gives the general idea and some examples, but it would not be the
main basis of an article for %2BBS%1, because I would want to cover
work by others as well as my own approach.

	The work on non-monotonic reasoning will also draw sparks - from
people with other opinions about how to do it than mine, from Bayesians
who have papered over some of the difficulties in finding suitable
sample spaces, and from philosophers who have supposed that %2ceteris
paribus%1 reasoning is necessarily informal.
.skip to column 1

	An article on non-monotonic reasoning would take longer to
prepare.  %2Ascribing Mental Qualities to Machines%1 would probably
attract more comment, since people are familiar with the issue, but
non-monotonic reasoning would introduce newer ideas.

	Which, if either, would be suitable?

.sgn