/LMAR=0/XLINE=3/FONT#0=BASL30/FONT#1=BASI30/FONT#2=BASB30/FONT#10=BAXM30/FONT#11=ZERO30/FONT#3=STA200/FONT#4=NGB25
␈↓ ↓H␈↓␈↓βS␈↓∧ Artificial Intelligence Laboratory, STANFORD UNIVERSITY, Stanford, California 94305

␈↓ ↓H␈↓∧Telephone 415 497-4430␈↓ 
⊗June 24, 1980 




␈↓ ↓H␈↓Dr. Stevan Harnad
␈↓ ↓H␈↓␈↓↓The Behavioral and Brain Sciences
␈↓ ↓H␈↓↓P.O. Box 777
␈↓ ↓H␈↓↓Princeton, New Jersey 08540

␈↓ ↓H␈↓Dear Dr. Harnad:

␈↓ ↓H␈↓        Thanks for your letter of June 17.  I have two possibilities for a ␈↓↓BBS␈↓ target article.

␈↓ ↓H␈↓        The␈α∪≡rst␈α∪is␈α∪based␈α∪on␈α∪a␈α∪paper␈α∪entitled␈α∪␈↓↓Ascribing␈α∪Mental␈α∪Qualities␈α∪to␈α∪Machines␈↓␈α∩which
␈↓ ↓H␈↓appeared␈αin␈αMartin␈α
Ringle's␈α␈↓↓Philosophical␈αPerspectives␈α
in␈αArti≡cial␈αIntelligence␈↓,␈αHumanities␈α
Press.
␈↓ ↓H␈↓The␈α
publisher␈α
seems␈α
to␈α
have␈α
done␈α
nothing␈α
with␈α
the␈α
book,␈α
and␈α
I␈α
don't␈α
even␈α
have␈α
a␈α
copy.␈α I␈α
enclose
␈↓ ↓H␈↓a␈α∩slightly␈α∩updated␈α∩version,␈α∩but␈α⊃I␈α∩would␈α∩want␈α∩to␈α∩update␈α⊃it␈α∩much␈α∩more␈α∩to␈α∩take␈α∩into␈α⊃account
␈↓ ↓H␈↓developments␈α
in␈α∞non-monotonic␈α
reasoning,␈α∞to␈α
take␈α
into␈α∞account␈α
philosophical␈α∞work␈α
I␈α∞have␈α
come
␈↓ ↓H␈↓across␈α
recently␈α(even␈α
though␈αsome␈α
of␈αit␈α
is␈α
old),␈αto␈α
correct␈αobscurities␈α
that␈αhave␈α
been␈α
pointed␈αout,
␈↓ ↓H␈↓and␈αto␈αgive␈αbetter␈αexamples.␈α It␈αalso␈αappeared␈αas␈αa␈αreport␈αof␈αour␈αlaboratory,␈αbut␈αthat␈αversion␈αwas
␈↓ ↓H␈↓based on the wrong computer ≡le and contains errors.

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

␈↓ ↓H␈↓        The␈αsecond␈αpossibility␈αis␈αa␈αreview␈αarticle␈αon␈αnon-monotonic␈αreasoning.␈α My␈αrecent␈αarticle␈α
on
␈↓ ↓H␈↓that␈αsubject␈αpresents␈αjust␈αone␈αapproach,␈αand␈αI␈αthink␈αthe␈α␈↓↓BBS␈↓␈αaudience␈αwould␈αbe␈αbetter␈αserved␈αby
␈↓ ↓H␈↓a comprehensive article.

␈↓ ↓H␈↓        Non-monotonic␈αreasoning␈αhas␈αthe␈αproperty␈αthat␈αa␈αconclusion␈α␈↓↓p␈↓␈αnon-monotonically␈αobtained
␈↓ ↓H␈↓from␈αa␈αcollection␈α␈↓↓A␈↓␈αof␈αpremisses␈αmay␈αnot␈αfollow␈αfrom␈αa␈αcollection␈α␈↓↓B␈↓␈αthat␈αcontains␈αall␈αthe␈αsentences
␈↓ ↓H␈↓of␈α∞␈↓↓A␈↓␈α∞and␈α∞more.␈α
 Logical␈α∞deduction␈α∞is␈α∞monotonic,␈α∞but␈α
humans␈α∞do␈α∞and␈α∞intelligent␈α∞machines␈α
must
␈↓ ↓H␈↓reason␈α∞non-monotonically␈α∂as␈α∞well.␈α∞ I␈α∂enclose␈α∞a␈α∂copy␈α∞of␈α∞a␈α∂report␈α∞based␈α∞on␈α∂a␈α∞paper␈α∂published␈α∞in
␈↓ ↓H␈↓␈↓↓Arti≡cial␈α
Intelligence␈↓,␈α
that␈α
gives␈α
the␈α
general␈α
idea␈αand␈α
some␈α
examples,␈α
but␈α
it␈α
would␈α
not␈α
be␈αthe␈α
main
␈↓ ↓H␈↓basis␈α∞of␈α∞an␈α∂article␈α∞for␈α∞␈↓↓BBS␈↓,␈α∞because␈α∂I␈α∞would␈α∞want␈α∞to␈α∂cover␈α∞work␈α∞by␈α∞others␈α∂as␈α∞well␈α∞as␈α∂my␈α∞own
␈↓ ↓H␈↓approach.

␈↓ ↓H␈↓        The␈α⊃work␈α⊂on␈α⊃non-monotonic␈α⊂reasoning␈α⊃will␈α⊂also␈α⊃draw␈α⊂sparks␈α⊃-␈α⊂from␈α⊃people␈α⊃with␈α⊂other
␈↓ ↓H␈↓opinions␈α∂about␈α∂how␈α⊂to␈α∂do␈α∂it␈α⊂than␈α∂mine,␈α∂from␈α⊂Bayesians␈α∂who␈α∂have␈α⊂papered␈α∂over␈α∂some␈α⊂of␈α∂the
␈↓ ↓H␈↓di≠culties␈α⊃in␈α⊃≡nding␈α⊃suitable␈α⊂sample␈α⊃spaces,␈α⊃and␈α⊃from␈α⊂philosophers␈α⊃who␈α⊃have␈α⊃supposed␈α⊂that
␈↓ ↓H␈↓␈↓↓ceteris paribus␈↓ reasoning is necessarily informal.
␈↓ ↓H␈↓αDr. Stevan Harnad␈↓ ¬qJune 24, 1980␈↓ 
nPage 2␈↓↓ 


␈↓ ↓H␈↓        An␈α⊂article␈α⊂on␈α∂non-monotonic␈α⊂reasoning␈α⊂would␈α⊂take␈α∂longer␈α⊂to␈α⊂prepare.␈α⊂ ␈↓↓Ascribing␈α∂Mental
␈↓ ↓H␈↓↓Qualities␈α∞to␈α∞Machines␈↓␈α∞would␈α∞probably␈α∞attract␈α∞more␈α∞comment,␈α∞since␈α∞people␈α∞are␈α∞familiar␈α∞with␈α∞the
␈↓ ↓H␈↓issue, but non-monotonic reasoning would introduce newer ideas.

␈↓ ↓H␈↓        Which, if either, would be suitable?


␈↓ ↓H␈↓Sincerely,



␈↓ ↓H␈↓John McCarthy
␈↓ ↓H␈↓Director
␈↓ ↓H␈↓Professor of Computer Science