Newsgroups: comp.ai.philosophy,sci.cognitive
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!lard.ftp.com!bloom-beacon.mit.edu!news!minsky
From: minsky@media.mit.edu (Marvin Minsky)
Subject: Re: Society of Mind
Message-ID: <1995May19.222112.9348@media.mit.edu>
Sender: news@media.mit.edu (USENET News System)
Cc: minsky
Organization: MIT Media Laboratory
References: <800772785snz@longley.demon.co.uk> <3pgjub$3ub@mp.cs.niu..edu> <800878008snz@longley.demon.co.uk>
Date: Fri, 19 May 1995 22:21:12 GMT
Lines: 54
Xref: glinda.oz.cs.cmu.edu comp.ai.philosophy:28196 sci.cognitive:7650

The attributions below are too complicated for me to unscramble.
In article <800878008snz@longley.demon.co.uk> David@longley.demon.co.uk writes:
>In>> 
>> >As I understand it, Cognitive Science is premised on a coherence thesis, ie
that individuals are consistent, or consistency seeking systems, which comes
down to a rationality assumption.  However, the evidence I  have  marshalled
suggests the opposite.

I don't think I've ever heard any cognitive scientists propose such a
thing.  Do you mean "philosopher"?
>> >In the 'Society of Mind', Minsky  develops a modular
>> >thesis which does not require the existence of a rational agent:

My SoM theory suggests that the brain has many parts, with specific
kinds of interconnections.  I would never use a term like "modular",
though, because it seems to suggest all sorts of things to different
people.  I do argue that many things that the mind does are done more
or less simultaneoulsy in several different ways, and this is why the
thing works so well, even when any particular part might produce
useless results.

>>  My impression, based on Minsky's occasional postings in
comp.ai.philosophy, is that he is not opposed to rational agency views.

I do oppose any idea that the mind proceeds in a generally
uniform, consistent manner, that can be described as rational in the
sense of "logical".  Worse, I don't know of any more sensible meaning
to the term "rational", and I'm pretty sure that I have never used it
or any synonym for it.

>> I don't read that as opposed to rational agency.  It is consistent with
>> SoM.  Note that they explicitly indicate differences with connectionism,
>> and it is connectionism which is often considered the alternative to
>> rational agency.

Perhaps the idea is that "symbol manipulating" systems usually produce
intermediate symbolic expressions that can sometimes be interpreted as
"reasons" for performing some of the steps.  In this sense "rational"
might be OK, if it means in the sense of "having reasons for".  Then
my reservation is simply that I don't thing that our "reasons" are
generally "valid", but only that (because of their evolutionary
development, both individually and historically) they tend to be heuristically
useful.

The research of Piaget and others generally show that people are
indeed neither inclined toward nor capable of "valid" logical
reasoning.  This, in my view, does little harm because we'd have no
reliable source of a sufficient supply of compact, unqualified, axioms
to which formal inference schemes could be applied.  Instead, I'd
argue, we "reason" primarily by matching and adapting many kinds of
patterns, by using different similarity-matching processes according
to our experience with which kinds work in various contexts.


