COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00003 00003	Minor
C00005 00004	How moods affect reasoning
C00009 00005		Haugeland has more or less convinced me that I have been somewhat
C00014 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source
Comments on John Haugeland's "The Nature and Plausibility of Cognitivism" 

	Haugeland's paper has convinced me that I am
indeed a %2cognitivist%1.  Naturally, I like to think that my
version of cognitivism is more subtle than what Haugeland describes.
Minor

Remarks about holograms.  If someone else says something similar,
please excise mine.

abstract - It is true that mental processes are not mainly described
by equations.  Neither is digestion or the habits of bees.  I.e.
cognitivism is not revolutionary in that respect.

5 I agree that %2systematic%1 explanations are required.

11 There is a continuum of reductions.  At one end we have the
reduction of thermodynamics to statistical mechanics.  If you know
statistical mechanics, then you can deduce all you want about
thermodynamics, and this deduction is reasonably practical.  In
the middle we have, say, the reduction of computers to structures
built of gates and latches.  You may know all about gates and latches,
but not know what you need to know about computers.  An extreme
example comes from the fact that numbers can be represented by
strings of beads, but he who knows all about beads may still be
puzzled by Fermat's last theorem.

12 The criterion for a device that plays chess can be stated more
rigorously.
How moods affect reasoning

	I agree with Haugeland that moods affect human reasoning in a way
unlike that of adjoining to a collection of sentences expressing one's
beliefs and observations additional sentences describing mood, and then
drawing logical conclusions from the augmented collection.  My explanation
invokes a mode of reasoning, called %2circumscription%1, that I have
proposed elsewhere (McCarthy 1977).

	It is often necessary to draw conclusion from insufficient
data, either because a person or program doesn't have all the data
or even because one's "mental capacity" is too small to take into
account all the data one has.  We propose that this is often done
by assuming that the entities whose existence is implied by the
facts one is taking into account are all the entities that exist
in certain categories.  For example, one may presume that the
advantages one has listed of a certain course of action are all
the advantages there are.  That these are all the advantages does
not follow from the individual sentences asserting that each of
them is an advantage.  We call %2circumscription%1 the mental
operation of presuming that an explicit list of objects exhausts
a class.  Various ways of formalizing it are discussed in the
reference.

	Circumscription is a powerful but unsafe mode of reasoning.
It is powerful in that it allows deciding what to do in cases
where a decision is necessary but information is insufficient.  It
is unsafe, since additional entities of the circumscribed classes
may actually exist.

	We propose that moods affect reasoning by affecting the class of
facts taken into account.
For example, a melancholy mood could directly retrieve facts
that have previously been classified as melancholy.  On the other
These sentences could in turn excite an information retrieval
mechanism to produce associated sentences.  This could even be
done by injecting a single %2I feel melancholy%1 sentence, although
this seems unlikely.
	Haugeland has more or less convinced me that I have been somewhat
of a cognitivist all my life, but in order to defend the doctrine, I need
to make it a little more sophisticated.

	We must first distinguish %2competence cognitivism%1 from
%2performance cognitivism%1.  The idea is that the output of
a mental process may be related to its inputs by a process that
can be described as drawing conclusions in some declarative formalism
without the actual mechanism by which the results were obtained
being like deduction in a formal system.
Arguments that the human brain obtains some of its results faster
than can be accounted for by deductive processes are relevant
criticisms of %2performance cognitivism%1 but not of %2competence
cognitivism%1.  I regard both kinds of cognitivism as having limited
application, but we might relate the output of many skilled actions
cognitivistically to the input without postulating a deductive internal
mechanism.

	Second, it isn't quite clear when a mechanism is to be considered
cognitive in character even in machines.  In particular, pattern
matching mechanisms are quite varied, and some of them are closer than
others to deductive reasoning.

	Concerning "understanding" I would like to make the following
conjecture.  A person can deal with symbolic inputs in either of two
ways.  First, he can manipulate it according to rules he has learned,
and second, he can translate it into his "internal language".  Only
in the latter case can he combine it freely with other information
to draw conclusions.  We say that someone has information but doesn't
understand it when he can manipulate the symbols as objects but hasn't
translated enough of it into his "internal language".  The main symptom
of lack of understanding, is a failure to draw certain conclusions
from symbolic information that seem "obvious" to those who do understand.
This phenomenon offers no problems to cognitivism of either kind, but
understanding it does require distinguishing symbolic expressions from
the information they express.

	The most difficult of Haugeland's challenges to cognitivism
is the problem of how moods affect reasoning.  I think the explanation
is related to mechanisms for jumping to conclusions from insufficient
data.  In (McCarthy 1978), I propose a mechanism called %2circumscription%1
for doing this, and for the present let us regard it as a %2competence%1
explanation, i.e. it gives the result of a mental process without
asserting that the human brain goes through the same steps.
%2Circumscribing%1 a collection of sentences is jumping to the conclusion
that the objects in a certain class whose existence is inferrable
from the collection of sentences exhaust the class.
In the paper, I argue that without such a mechanism, it would hardly
ever be possible for a human or machine to draw conclusions from
the data that is actually available in common sense situations.

	Let us now imagine moods to be chemical, e.g. a melancholy
mood is just a high concentration of %2melancholine%1 in the blood.