/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BAXB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRK30



␈↓ ↓H␈↓Comments on John Haugeland's "The Nature and Plausibility of Cognitivism"

␈↓ ↓H␈↓␈↓ α_Haugeland's␈α∂description␈α∂of␈α∂the␈α∂attractions␈α⊂of␈α∂cognitivism␈α∂convinced␈α∂me␈α∂that␈α∂I␈α⊂have␈α∂long
␈↓ ↓H␈↓been␈α∃a␈α∀cognitivist,␈α∃although␈α∃I␈α∀would␈α∃not␈α∃try␈α∀to␈α∃give␈α∃cognitive␈α∀explanations␈α∃to␈α∃all␈α∀mental
␈↓ ↓H␈↓phenomena.␈α
 His␈α
challenges␈α
to␈α
cognitivism␈α
are␈αinteresting,␈α
and␈α
here␈α
are␈α
some␈α
tentative␈α
ideas␈αfor
␈↓ ↓H␈↓meeting them.

␈↓ ↓H␈↓␈↓ α_We␈α∪first␈α∀distinguish␈α∪␈↓↓competence␈α∪cognitivism␈↓␈α∀from␈α∪␈↓↓performance␈α∀cognitivism␈↓.␈α∪ Competence
␈↓ ↓H␈↓cognitivism␈α⊃is␈α∩tries␈α⊃to␈α⊃relate␈α∩the␈α⊃output␈α∩of␈α⊃a␈α⊃mental␈α∩process␈α⊃to␈α∩its␈α⊃inputs␈α⊃as␈α∩the␈α⊃result␈α∩of␈α⊃a
␈↓ ↓H␈↓deduction␈αto␈αits␈αpremises.␈α Performance␈αcognitivism␈αregards␈αthe␈αactual␈αmechanism␈αas␈αa␈αprocess␈αof
␈↓ ↓H␈↓deduction.␈α
 I␈α
would␈αsupport␈α
performance␈α
cognitivism␈α
for␈αsome␈α
mental␈α
phenomena␈αand␈α
competence
␈↓ ↓H␈↓cognitivism␈α∞to␈α∂many␈α∞more.␈α∂ Arguments␈α∞that␈α∂the␈α∞human␈α∞brain␈α∂obtains␈α∞some␈α∂of␈α∞its␈α∂results␈α∞faster
␈↓ ↓H␈↓than␈α
can␈αbe␈α
accounted␈αfor␈α
by␈αdeductive␈α
processes␈αare␈α
proper␈αchallenges␈α
to␈α␈↓↓performance␈α
cognitivism␈↓
␈↓ ↓H␈↓but␈α↔not␈α_to␈α↔␈↓↓competence␈α_cognitivism␈↓.␈α↔ Thus␈α_we␈α↔may␈α↔relate␈α_the␈α↔output␈α_of␈α↔a␈α_skilled␈α↔action
␈↓ ↓H␈↓cognitivistically to its input without postulating a deductive internal mechanism.

␈↓ ↓H␈↓␈↓ α_Performance␈α
cognitivism␈α
needs␈αmore␈α
precision,␈α
because␈αit␈α
isn't␈α
clear␈αwhen␈α
a␈α
mechanism␈αis␈α
to
␈↓ ↓H␈↓be␈αconsidered␈αcognitive␈α
even␈αin␈αmachines.␈α In␈α
particular,␈αpattern␈αmatching␈αmechanisms␈α
are␈αquite
␈↓ ↓H␈↓varied, and some of them are closer than others to deductive reasoning.

␈↓ ↓H␈↓␈↓ α_Here␈α∞is␈α
a␈α∞conjecture␈α∞about␈α
"understanding".␈α∞ A␈α∞person␈α
can␈α∞deal␈α∞with␈α
a␈α∞symbolic␈α∞input␈α
in
␈↓ ↓H␈↓either␈αof␈αtwo␈αways.␈α First,␈αhe␈αcan␈αmanipulate␈αit␈αaccording␈αto␈αrules␈αhe␈αhas␈αlearned,␈αand␈αsecond,␈αhe
␈↓ ↓H␈↓can␈αtranslate␈αit␈αinto␈αhis␈α"internal␈αlanguage".␈α Only␈αin␈αthe␈αlatter␈αcase␈αcan␈αhe␈αcombine␈αit␈αfreely␈αwith
␈↓ ↓H␈↓other␈α∀information␈α∀to␈α∀draw␈α∀conclusions.␈α∀ We␈α∀say␈α∀that␈α∀someone␈α∀has␈α∀information␈α∀but␈α∀doesn't
␈↓ ↓H␈↓understand␈α
it␈α
when␈αhe␈α
can␈α
manipulate␈α
the␈αsymbols␈α
as␈α
objects␈α
but␈αhasn't␈α
translated␈α
enough␈α
of␈αit
␈↓ ↓H␈↓into␈α
his␈α∞"internal␈α
language".␈α∞ The␈α
main␈α∞symptom␈α
of␈α∞lack␈α
of␈α∞understanding,␈α
is␈α∞a␈α
failure␈α∞to␈α
draw
␈↓ ↓H␈↓certain␈α∞conclusions␈α
from␈α∞symbolic␈α
information␈α∞that␈α
seem␈α∞"obvious"␈α
to␈α∞those␈α
who␈α∞do␈α
understand.
␈↓ ↓H␈↓This␈α⊃phenomenon␈α∩offers␈α⊃no␈α⊃problems␈α∩to␈α⊃cognitivism␈α⊃of␈α∩either␈α⊃kind␈α⊃provided␈α∩we␈α⊃distinguish
␈↓ ↓H␈↓symbolic expressions from the information they express.

␈↓ ↓H␈↓␈↓ α_The␈α
most␈α
difficult␈α
of␈α
Haugeland's␈α
challenges␈α
to␈α
cognitivism␈α
is␈α
the␈α
problem␈α
of␈α∞how␈α
moods
␈↓ ↓H␈↓affect␈αreasoning.␈α It␈αis␈αmost␈αacute␈αif␈αwe␈αimagine␈αmoods␈αto␈αbe␈αchemical;␈αe.g.␈αa␈αmelancholy␈α
mood␈αis
␈↓ ↓H␈↓just␈α
a␈α
high␈α
concentration␈α
of␈α
␈↓↓melancholine␈↓␈α
in␈α
the␈α
blood␈α
-␈α
together␈α
with␈α
its␈α
effects.␈α
 How␈α
can␈α
this
␈↓ ↓H␈↓affect␈α∞what␈α
long␈α∞term␈α∞goals␈α
seem␈α∞worth␈α∞pursuing␈α
-␈α∞as␈α
moods␈α∞often␈α∞do?␈α
 My␈α∞proposals␈α∞here␈α
are
␈↓ ↓H␈↓tentative and certainly require revision.

␈↓ ↓H␈↓␈↓ α_The␈α∃presence␈α∃of␈α∃␈↓↓melancholine␈↓␈α∃might␈α∃cause␈α∀certain␈α∃sources␈α∃of␈α∃sentences␈α∃to␈α∃come␈α∀into
␈↓ ↓H␈↓consciousness␈α
with␈α
higher␈α
probability,␈α∞or␈α
to␈α
present␈α
themselves␈α∞with␈α
greater␈α
urgency,␈α
or␈α∞to␈α
come
␈↓ ↓H␈↓with␈αgreater␈αfrequency.␈α This␈αassumes␈αthat␈αthe␈αinformation␈αretrieval␈αsystem␈αthat␈αdetermines␈αwhat
␈↓ ↓H␈↓comes into consciousness is partly non-logical.  Likewise it inhibits other sources of sentences.

␈↓ ↓H␈↓␈↓ α_The␈α
presence␈α
of␈α
␈↓↓melancholine␈↓␈αmight␈α
give␈α
rise␈α
to␈α
a␈αsentence␈α
␈↓↓I␈α
am␈α
melancholy␈↓.␈α
 Perhaps␈αthis
␈↓ ↓H␈↓sentence has no special effect by itself, but can trigger learned effects.

␈↓ ↓H␈↓␈↓ α_The␈α∀sentence␈α∪␈↓↓"I␈α∀am␈α∪melancholy"␈↓␈α∀might␈α∀be␈α∪used␈α∀logically␈α∪by␈α∀the␈α∀information␈α∪retrieval
␈↓ ↓H␈↓mechanism.␈α
 Thus␈α
there␈α
might␈α
be␈αrules␈α
saying␈α
something␈α
like␈α
␈↓↓"If␈αI␈α
am␈α
melancholy,␈α
I␈α
should␈αlist
␈↓ ↓H␈↓↓things that might go wrong"␈↓.  This might work but seems subjectively implausible.
␈↓ ↓H␈↓␈↓ εH␈↓ 91


␈↓ ↓H␈↓␈↓ α_All␈αthese␈αhypotheses␈αrequire␈αan␈αexplanation␈αof␈αa␈αphenomenon␈αwhich␈αoccurs␈αin␈αmany␈αother
␈↓ ↓H␈↓contexts.␈α
 A␈α
certain␈α
set␈α␈↓↓A␈↓␈α
of␈α
sentences␈α
can␈αlead␈α
to␈α
a␈α
conclusion␈α
␈↓↓p.␈↓␈α Another␈α
set␈α
␈↓↓B␈↓␈α
consistent␈αwith␈α
␈↓↓A␈↓
␈↓ ↓H␈↓and␈αmaybe␈αeven␈αcontaining␈α␈↓↓A␈↓␈αcan␈αlead␈αto␈αthe␈αconclusion␈α␈↓↓¬p.␈↓␈α(McCarthy␈α1978)␈αascribes␈αthis␈α
to␈αa
␈↓ ↓H␈↓method␈α!of␈α!conjectural␈α reaoning␈α!we␈α!call␈α ␈↓↓circumscription␈α!induction␈↓.␈α! Briefly,␈α applying
␈↓ ↓H␈↓circumscription␈α⊂induction␈α⊂involves␈α⊂jumping␈α⊂to␈α⊂the␈α⊂conclusion␈α⊂that␈α⊂the␈α⊂objects␈α⊂whose␈α⊂existence
␈↓ ↓H␈↓follows␈αfrom␈αa␈αset␈α
␈↓↓A␈↓␈αof␈αsentences␈αare␈α
all␈αthe␈αobjects␈αthere␈αare␈α
in␈αa␈αcertain␈αclass.␈α
 This␈αconclusion
␈↓ ↓H␈↓may␈αlead␈αin␈αturn␈αto␈αthe␈αconclusion␈α␈↓↓p.␈↓␈α Enlarging␈α␈↓↓A␈↓␈αto␈α␈↓↓B␈↓␈αmay␈αbring␈αin␈αnew␈αobjects,␈αand␈αthis␈αmay
␈↓ ↓H␈↓lead␈α
to␈α
␈↓↓¬p.␈↓␈α
 Without␈α
some␈α
such␈αmechanism␈α
we␈α
wouldn't␈α
be␈α
able␈α
to␈α
draw␈αcontradictory␈α
conclusions
␈↓ ↓H␈↓from consistent sets of information merely by neglecting part of it.

␈↓ ↓H␈↓Reference:

␈↓ ↓H␈↓␈↓αMcCarthy,␈αJohn␈↓␈α(1978),␈α␈↓↓Circumscription␈αInduction␈α-␈αA␈αWay␈α of␈αJumping␈αto␈αConclusions␈↓,␈αStanford
␈↓ ↓H␈↓Artificial Intelligence Laboratory, (to be published).

␈↓ ↓H␈↓John McCarthy
␈↓ ↓H␈↓Artificial Intelligence Laboratory
␈↓ ↓H␈↓Stanford University
␈↓ ↓H␈↓Stanford, California 94305