Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 193
AIList Digest Monday, 3 Aug 1987 Volume 5 : Issue 193
Today's Topics:
Philosophy - Natural Kinds
----------------------------------------------------------------------
Date: 30 Jul 87 16:25:29 GMT
From: aweinste@bbn.com (Anders Weinstein)
Reply-to: aweinste@bbn.com (Anders Weinstein)
Subject: Re: Natural Kinds (Re: AIList Digest V5 #186)
In article <MINSKY.12321721233.BABYL@MIT-OZ> MINSKY@OZ.AI.MIT.EDU writes:
>
>My conclusion - and, I'd bet, Ken Laws would agree - is that the
>concept of "natural kind" has an illusory generality. It seems to me
>that, rather than good philosophy, it is merely low-grade science
>contaminated by naive, traditional common sense concepts.
I think there's some confusion about what natural kinds are in this
discussion. Most of the talk has focussed on the alleged sharpness of the
kind's boundaries. But I don't think this is what's at issue, at least in the
contemporary philosophical usage.
The point is that you can't do science without imposing some taxonomy on the
objects under study. "Natural kinds" are simply the kinds that figure in
scientific generalizations (aka Laws of Nature). Thus "bird" is perhaps a
natural kind, but "thing that is either furry or made of clay" is not.
Some people like to argue about whether these classification systems are "out
there" in Nature waiting to be discovered (the "realist" view) or are
invented by the mind and imposed on some undifferentiated reality (an
"idealist" or "constructivist" picture). Happily, we can ignore this debate.
What we can't ignore is the fact that a notion of natural kinds is
*essential* for induction, as demonstrated by Nelson Goodman's classic "grue
vs green" puzzle. Without some sense of what kinds are "natural", you're
liable to go off projecting "grue", or looking for laws governing "furry or
clay things". This would be the antithesis of intelligence.
Of course, coming up with suitable taxonomies is an empirical matter. I once
heard Kuhn emphasize that Aristotle's concept of "motion" included things
like the growth of trees. Progress in physics had to await a more useful
concept of motion.
But this shouldn't be taken to imply that natural kinds are only relevant to
sophisticated scientific theorizing -- the same principles apply to the
inductions that are part of common-sense understanding. And it seems that we
are blessed with pretty accurate innate intuitions about which kinds or
similarities are natural (eg. "green") and which are ludicrously artificial
("grue"). The philosophy of induction thus suggests that you can't make an
intelligent system without somehow building into it an equivalent sense of
the naturalness of kinds.
Anders Weinstein
BBN Labs
------------------------------
Date: Thu, 30 Jul 87 11:39:00 n
From: Paul Davis <DAVIS%EMBL.BITNET@wiscvm.wisc.edu>
Subject: Natural Kinds
As the list of different aspects important in the recognition
of a natural kind grows, I'd like to throw in a comment inspired by
the mention of Balans chairs, amongst other things.
My experience of Balans chairs, apart from extreme comfort, is
that peoples recognition of them tend to be via contextual means
rather than anything else - ie; "that thing in front of the desk must
be a chair" (implicitly - "chairs occur in front of desks and that
thing looks sufficiently chair-like, and insufficiently
anything-else-that-I've-ever-seen-like, to be an example of a
chair"). Interestingly though, people do not initially know how to
*use* the chair (unless they've used one before or seen it in use).
This suggests to me that contextual information is at least as
important as structural or functional types in identifying a `natural
kind'.
To support this a little further, I have my own experience of
living in a country where I have a very limited grasp of the language.
Although clearly language understanding could be argued to be a
different case to that of natural kinds, I take my ability to deduce
the 'kind' of a notice or message written in German from a very
limited vocabulary but a very large `database' (yuk) of contexts to
indicate, perhaps, the kind of thing thats going on. This ability
arises because I know from past experience what kinds of signs and
messages occur in what kind of contexts, and with a few known words, a
few vague resemblances between German and English, its pretty easy to
figure out a reasonable guess at the meaning.
Both of these examples seem to me to point towards a more
general recognition of the Dreyfus's theory of skill acquisition.
Herbert and Stuart Dreyfus argue that contrary to many (most ?)
current notions of skill acquisition, the move from novice (baby ?) to
expert (adult ?) proceeds by moving from abstract understanding to a
position reliant on having a large repetoire of specific experiences.
Why should the recognition of `natural kinds' differ from
this ? So far, most commentators have been focusing on an attempt to
arrive at an abstract definition of a natural kind. Why should this reflect
reality better than the alternative - that we recognise a natural
kind, or anything else for that matter, simply by having sufficient
past experience to do so. Of course, the definition-by-recognition of
a natural kind will actually be a consequence of the interaction of
past experiences and a pattern recognition ability, but I can see no
reason to believe that the classification of natural kinds is any
different to the potentially more general question of how do we
recognise similarity. Natural kinds differ only in that they represent
the ability to group a large number of experiences rather than just
two or three.
Why do I get the feeling that I've missed the point ?
Paul Davis
Biocomputing, EMBL, Postfach 10.2209, 6900 Heidelberg, West Germany
bitnet/earn: davis@embl.bitnet
arpa: davis%embl.bitnet@wiscvm.wisc.edu
uucp: ...!psuvax1!embl.bitnet!davis
------------------------------
Date: 30 Jul 87 18:30:32 GMT
From: aweinste@bbn.com (Anders Weinstein)
Reply-to: aweinste@bbn.com (Anders Weinstein)
Subject: Re: Natural Kinds (The Putnam/Kripke analysis)
In an earlier posting, McCarthy referred to Putnam's work on natural kinds. I
think what is important about this work is not the mere notion of a "natural
kind," which is ancient, but rather the new *analysis* of the semantics of
natural kind terms made by Putnam and Kripke. And, as McCarthy suggested,
this analysis may well be relevant to terms other than scientific kind
predicates.
In brief, the important point is that the marks by which we identify members
of a kind are *not* in general definitive. Gold may be identified as a yellow
metal, but even if we suppose that this description *uniquely* refers to
gold, still "gold" is not synonymous with "yellow metal". Put another way, it
is not a necessary truth that gold is yellow or that gold is a metal: it
could conceivably turn out that we were mistaken on either of these points.
By extending such examples, Putnam and Kripke make the claim is that there is
initially *no* criterial definition of gold; neither a simple definition nor
some weighted disjunctive cluster of properties will do. On the causal
history view of reference, a statement like "gold is F" is not semantically
equivalent to any assertion about "a yellow metal". It is rather a little bit
like waving a rope in the air and saying "the stuff on the end of this rope
is F", where the "rope" is a chain of reference-preserving causal links
stretching back in time to the introduction of the term.
A couple of other points in Putnam's analysis are also relevant. First, he
claims that although definitive criteria for natural kind terms are not
initially known, they may be uncovered in time by scientific investigation.
Thus, today we believe we have found the "essence" of gold, namely being an
element with a certain atomic number. This is information the Greeks didn't
have; nevertheless, we generally regard our word "gold" as having the same
meaning as the ancients' word for it.
But even after a criterial definition is discovered, still there is a
"division of linguistic labor" -- the precise definition may only be known by
experts. Lay members of the language community are judged to be competent
users of the concept as long as they possess certain (non-definitive)
stereotypical information about the kind in question. Knowing that an elm is
some kind of deciduous tree is enough to be credited with understanding
"elm", even if you can't distinguish elms from beeches or any other trees.
Well, what is the significance of all this for cognitive science? At first,
it seems to demonstrate that meanings are partly extra-psychological, because
of the role of the causal history in fixing the reference of a term. As
Putnam puts it, "meanings just ain't in the head." And this might suggest
that apart from the demand for stereotypes, Putnam's points are just
irrelevant to the psychology of concept understanding, insofar as psychology
is limited to dealing with what *is* in the head.
Nevertheless, the theory does have *some* implications for psychological
theories. Broadly speaking, it implies that a cognitive system must be
prepared to deal appropriately with the phenomena that Putnam describes. That
is, it must "understand" that the marks by which it identifies things need
not constitute a *definition*, so it can make sense of the possibility of,
say, a blue lemon. This suggests that there is more to knowing the meaning
of a term than merely having an inner criterion of application -- one must
also be able to make sense of situations where the criterion fails.
Also, the system must understand that relevant experts may know more about
the term than it does, and that a truly essential criterion may yet be
discovered. In a good paper on this subject by Georges Rey (I will dig up the
citation when I get a chance), Rey puts the point in AI-style language: the
representation of a concept must have a "slot" for the definitive criterion
which may initially be unfilled. Although this formulation seems to me to
involve a vast (but typical) over-simplification of understanding, it does
indicate how the Putnam analysis is relevant to cognitive theories.
Anders Weinstein
BBN Labs
------------------------------
Date: 30 Jul 87 18:49:24 GMT
From: aweinste@bbn.com (Anders Weinstein)
Reply-to: aweinste@bbn.com (Anders Weinstein)
Subject: Natural Kinds (The Putnam/Kripke analysis)
BTW, I just reread McCarthy's original posting on kinds and it seems that the
consequences of the Putnam analysis that I gave are substantially the same as
the points McCarthy was making. However, subsequent respondants seem to have
concentrated exclusively on the supposed sharpness of kind boundaries.
Although this is alluded to in McCarthy's message, I don't think it's germane
to the main points about concept understanding and definitions.
Anders Weinstein
------------------------------
Date: Thu, 30 Jul 1987 22:43 EDT
From: MINSKY%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU
Subject: AIList Digest V5 #190 - Msc.
Stan Shebs asks "how *can* these hypotheses be tested?" and suggests
that many of the ideas in Society of Mind follow human thinking rather
than minds in general. Yes, I focussed on trying to make a large
scale theory of human psychology. As for testing it, "implementation"
in a computer model is one possibility, but for the near future at
least, as Stan says, "that would probably be handicapped by being too
small and simple to be recognizably human-like in its behavior."
My answer is simple and direct: we must work toward building
instruments to help us see what is happening in the brain! In a few
decades I hope we shall see, for example, SQUID quantum-based magnetic
scanners that can map current flow, at least in the cerebral cortex,
to a resolution of better than a millimeter. That will tell us a
great deal about activities during thinking. Ultimately, we need
probes that smap down to better than that, at least in a significant
region. My point is that we ought not assume that we shall always be
limited to the crude psychological-response methods presently
available, or low resolution brain-wave or PETT-scan instruments.
------------------------------
Date: 31 Jul 87 20:48:51 GMT
From: aweinste@bbn.com (Anders Weinstein)
Reply-to: aweinste@bbn.com (Anders Weinstein)
Subject: Reference on concepts & natural kinds
For those interested in the implications of the Putnam/Kripke philosophy of
natural kinds for the psychology of concepts, I recommend the article I
alluded to in a previous message, Georges Rey's "Concepts and Stereotypes",
Cognition 15:1-3 (1983).
This piece is a philosophically informed critique of Smith & Medin's book
"Categories and Concepts" (1981). Medin & Smith offer a reply in Cognition
17:3 (1984), and Rey answers their reply in Cognition 19:3 (1985).
Naturally the bibliographies of these articles point to many other relevant
papers; I won't list them here.
Anders Weinstein
BBN Laboratories
------------------------------
End of AIList Digest
********************
