Re: Kid-sib and computation

From: Harnad, Stevan (
Date: Sat Feb 22 1997 - 22:52:29 GMT

> From: Cherry, Sandra <>
> This is what I have got from computation; it is a way of modeling
> (producing computer programs) as to what is going on in our minds. It
> is what we do "rationally" - but, it is done unconsciously- i.e. we are
> unaware of it and therefore Cogsci seeks to eplain these properties by
> way of computer models.

So far you are right. Computation actually plays two roles:

(1) It is used to test theories about the mind; when computers are used
that way (whether for cognitive modeling or for astronomic modeling)
then they are just fast substitutes for paper and pencil. This way of
using computers does not imply that cognition is computation, any more
than it implies that planets compute their orbits.

(2) In computationalIST modeling, the hypothesis is that the
computations that the computer is doing are also done by our brains.

I, for example, think (1) is ok, but (2) is wrong (because of the symbol
grounding problem).

> Physical appearance is irrelevant, it is the
> shapes that are important.

Not clear what you are saying here. And what is the difference between
physical appearance and shape?

I think you mean that the shape of symbols is arbitrary: they do not
look like the things they refer to and they are not phsically connected
to the things they refer to, any more than the words in a book resemble
or are physically connected the person they describe. The symbol
manipulation rules apply to these shapes (take 0010 and adde 1101),
not their meanings.

With analog processing, in contrast, the resemblance of the image to
the thing it's an image of plays a crucial role: In matching rotated
and unrotated shapes, the analg representation is manipulated: it is
rotated to match the shape the subject is seeing.

In neither symbol manipulation nor image manipulation are we
necessarily aware of he process; it may be unconscious, hence
implicit. Symbolic processes are more likely to be explicit, but
they need not be.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:50 GMT