From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!ccu.umanitoba.ca!access.usask.ca!kakwa.ucs.ualberta.ca!alberta!aunro!ukma!wupost!darwin.sura.net!jvnc.net!nuscc!hilbert!smoliar Thu Feb 20 15:21:37 EST 1992
Article 3817 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca comp.ai.philosophy:3817 sci.philosophy.tech:2154
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!ccu.umanitoba.ca!access.usask.ca!kakwa.ucs.ualberta.ca!alberta!aunro!ukma!wupost!darwin.sura.net!jvnc.net!nuscc!hilbert!smoliar
>From: smoliar@hilbert.iss.nus.sg (stephen smoliar)
Newsgroups: comp.ai.philosophy,sci.philosophy.tech
Subject: Re: Robotic Follies
Message-ID: <1992Feb18.000808.12473@nuscc.nus.sg>
Date: 18 Feb 92 00:08:08 GMT
References: <1992Feb10.123736.8691@husc3.harvard.edu> <1992Feb15.011214.24421@nuscc.nus.sg> <1992Feb16.211812.8890@husc3.harvard.edu>
Sender: usenet@nuscc.nus.sg
Reply-To: smoliar@iss.nus.sg (stephen smoliar)
Organization: Institute of Systems Science, NUS, Singapore
Lines: 148

In article <1992Feb16.211812.8890@husc3.harvard.edu> zeleny@boucher.harvard.edu
(Mikhail Zeleny) writes:
>In article <1992Feb15.011214.24421@nuscc.nus.sg> 
>smoliar@iss.nus.sg (stephen smoliar) writes:
>
>>  It seems
>>to me that Minsky is questioning whether or not the treatment of beliefs as
>>propositional attitudes is an adequate model for some of the questions we may
>>with to ask about the behavior of a reasoning individual.  
>
>Hoping to avoid the repetition of the erstwhile inflammatory exchanges on
>the subject of "informal logic", I am nonetheless compelled to inquire:
>What do any of the "questions we may wish to ask about the behavior of a
>reasoning individual" have to do with the subject matter of a discipline
>ostensibly dedicated to the study of *valid* forms of reasoning?  If there
>be doxastic logic, it must concern itself with questions of valid reasoning
>about beliefs, no more, no less.  The study of behavior belongs to the
>purview of natural sciences; likewise for the study of practical reasoning. 
>
I, too, do not wish to stir up those coals of "informal logic" which offered
far more heat than light.  Let me first propose which I feel may be a more
suitable choice of my own words:  "questions we may with to ask about the
behavior of AN INDIVIDUAL MANAGING SUCCESSFULLY IN ITS ENVIRONMENT."  In
other words I would argue that Minsky is not necessarily concerned with
the various ways in which we may wish to interpret the word "reasoning;"
his primary concern is just with how we manage at all.  Now, I do not claim
to speak for the entire artificial intelligence community, nor, do I believe,
does Minsky.  Indeed, the community is so diverse that anyone claiming to do
so would be presumptuously arrogant.  If you choose to view artificial
intelligence as "a discipline ostensibly dedicated to the study of *valid*
forms of reasoning," then it would be a mistake to assume that your own view
is shared throughout the artificial intelligence community.  Some of us are
genuinely interested in how it is that vast numbers of agents out there are
capable of getting on in the world at all, even when large portions of their
behavior are blatantly irrational by any formal standard.  I can only speak
for myself, of course;  but I have a strong suspicion that Minsky shares my
interest.  Now I have no trouble at all with your categorizing such pursuits
under "the purview of natural sciences."  To a great extent, this is what
Minsky has been saying all along (although perhaps he should have been more
explicit).  His approach to artificial intelligence is, indeed, grounded on
the study of behavior;  and, assuming that grounding, he is basically arguing
that propositional attitudes do not provide an adequate model for addressing
the questions which lie at the heart of that study.  This is not so much an
attack on formal logic as it is an observation that you do not use a
screwdriver to cut off pieces of a two-by-four.

>>					       We start with an observer O who
>>is observing some entity E.  E' is said to be a MODEL of E if O can use E' to
>>answer questions that he has about E.  Note that Minsky does not say anything
>>specific about WHAT those questions are.  What is important is the ternary
>>relation which unites the observer's questions, the entity, and its model.)
>>Thus, Minsky is interested in questions such as, "Why does this child keep
>>repeating the action of piling blocks in a tower and then knocking them
>>down?"
>>or "Why does the child walk away from the blocks and look for something else
>>to
>>do?"  Minsky is questioning the adequacy of propositional attitudes as a
>>foundation for dealing with such psychological questions of mind (since
>>presumably the child is making some conscious decisions here);  and I,
>>personally, do not think that his doubts have yet been satisfactorily
>>addressed.
>
>Again, what does all of this have to do with logic of belief?

Again, all of this has NOTHING to do with logic of belief;  so let us see why
you think this deserves criticism.

>  If we are to
>believe Freud (and Minsky certainly does), the reasons we may adduce for
>our actions are nothing but ex post factum rationalizations of subconscious
>conative impulses.  Setting this Schopenhaur-flavored Weltanschauung aside,
>we may try to answer Minsky's questions with a full-blown theory of behavior
>dispositions; however, unless you are prepared to be a behaviorist about
>belief, this course of action would be orthogonal to our issue.
>
I think you have now set yourself (and probably many of critical members of the
artificial intelligence community) an excellent agenda for reading THE SOCIETY
OF MIND.  This is not to say that the book builds of that "full-blown theory of
behavior dispositions" step by step (or even in fits and starts).  However, it
starts you down the road to building such a theory;  and, yes, going down that
road IS orthogonal to those issues of logic of belief which you have studied so
intensely.  I would, however, may one suggestion in terms of situating this
approach with respect to its intellectual predecessors.  We both agree on the
debt to Freud, but the sort of behaviorism which Minsky is pursuing is not
really the behaviorism of Skinner.  A much more significant influence is THE
STUDY OF INSTINCT by Nikolaas Tinbergen.  Minsky is very interested in the idea
that instinct (which, perhaps, may also be associated with Freudian drives) is
responsible for entire PATTERNS of behavior;  and he views those patterns as
products of entities very much like the agents of his society.  This leads
to two fundamental questions:

	1.  Where do those patterns come from?

	2.  If you have large numbers of them, how are they controlled?

This is ostensibly a rather more sophisticated view of behavior than can be
accommodated by ringing changes on the stimulus-response loop.  Now THE SOCIETY
OF MIND does not provide pat answers to these questions, and one may even agree
with critics who would argue that it does not provide ANY answers to them.
However, I would argue that it is a valid alternative to formal logic as a
focus of attention if we are basically interested in getting on the world
(and, perhaps, building some robots which can do so).

>SS:
>>One reason for doubting the power of propositional attitudes involves their
>>own foundation on terms which denote concepts.  In THE MISMEASURE OF MAN
>>(just
>>to shift the arena over to Harvard, Mikhail), Stephen Jay Gould discusses the
>>fallacy of REIFICATION--to wit, the misguided assumption that any word (term)
>>we use can be reified into a concept.  Interestingly enough, the particular
>>word Gould is picking on in this book is "intelligence."  Now, clearly, we
>>can do all sorts of interesting things with these terms at a strictly formal
>>level;  but to assume that those formal results may serve as a model of the
>>human (or any animal) behavior we observe may lead to an inadvertent
>>commission
>>of this reification fallacy.  It seems to me that the burden of proof lies
>>with
>>the builder of the model being able to convince the observer that his
>>questions
>>can now be answered satisfactorily.
>
>All observation that concerns me in this case is given by the data of
>descriptive linguists.  My interest is in providing a theory of linguistic
>meaning; for well-known reasons, it turns out that this theory is possessed
>of a metaphysical import.  Hence we obtain an imposing abstract ontology of
>propositions and concepts.  As for Gould's worries about reification, they
>concern me no more than Jerry Fallwell's creationist theories concern him;
>to put it bluntly, as a philosopher, Gould makes a damn good journalist.
>
Of course, Gould is not TRYING to be a philosopher in this book;  so it makes
little sense to criticize him for not being one!  He IS trying to be a natural
scientist (to use your own terminology);  and I think he is succeeding at it
rather well.  He is concerned with how we make observations and how we may
misinterpret the observations we make.  (He even includes an anecdote about
his OWN misinterpretation while analyzing the data discussed in this book,
lest you think he is trying to take a God's-eye view of natural science.)
On the basis of his studies, he proposes, as a conclusion, that the word
"intelligence" cannot be reified into a concept which is susceptible to
measurement.  For you this may move him into philosophy.  For me it is
a result based on interpretation of evidence which is up there along with
the observation that position and momentum of electrons cannot both be
determined with exact precision.  Just let Gould and Minsky go about with
their natural science, and let's call a peace!
-- 
Stephen W. Smoliar; Institute of Systems Science
National University of Singapore; Heng Mui Keng Terrace
Kent Ridge, SINGAPORE 0511
Internet:  smoliar@iss.nus.sg


