Comments on Putnam's "Computational Psychology and Interpretation Theory

1. I think I agree with Putnam that cognitive psychology does not require
a notion of %2sameness of content%1 for mental representations.  I would
go much further.  In fact I favor giving up notions of sameness, or at
least making them  secondary, of almost all concepts.  
This h`s two virtues:

	a. It makes possible systematic ambiguity between terms representing
objects and terms representing concepts.  This seems to be true of human
languages, and it may avoid misplaced specificity in cmputerl mentalese.

	b. It will make approximate and partial theories more feasible.

Most likely, neither of these reasons corresponds to what Putnam has
in mind.  

2. p.1 - Putnam is over generous to both depth psychology and 
behaviorism.
Neither is well defined enough to be simulated by a computer program.
Skinner would have to improvise in order to answer the questions of
the programmer and even then would get into a blind alley and would
have to start over.  After several tries, he would blame the programmer
and give up.

3. p.2 - What Putnam says about experience in programming computers
ought to be true, and I think it will, but, alas, programming reasoning
hasn't been well done yet.

4. p.3 - Putnam has identified the "working hypothesis of cognitive
psychology" with what I would like it to be.  Few cognitive psychologists
understand or take seriously expressing what people know in a
logical language or what they can infer as its logical consequences.
Part of the problem is that non-monotonic reasoning is required, and
its formal treatment has just begun.

	I should ask Zenon's opinion of this characterization.

5. p.5 - I incline to the view that mentalese doesn't have semantics
at all, let alone a verificationist semantics.  When a child hears
about a unicorn, it does not thereby acquire a test for whether
an object is a unicorn.  It merely accepts the idea that "unicorn"
designates something - real or mythical - and it is prepared 
learn more about unicorns should the occasion arise.

	Rather than no semantics at all, this suggests an
intensional semantics.   Intensional semantics can be carried
out in a Tarskian way by introducing concepts as objecsts as
in (McCarthy 1979).

6. p.6 - No we don't need a utility function, because the role
of language in thought can be substantially explained with just
the Missouri program and without a reasoning program per se -
let alone a goal-seeking program.

7. p.14 - I agree that interpretation is a holistic enterprise, but
partial and approximate interpretations are needed.

8. p.17 - While restriction of domain leads to one type of approximate
interpretation, it seems to me that there is another kind of
approximate interpretation, somehow dual to the first, that does
not involve restricting the domain but whose results are not
unrestrictedly correct.  We may either say more and more about
less and less or vice versa.

9. p.19 - I am surprised by this interpretation of Tversky.  There
is no reason to suppose that even genuine preferences are transitive.
Given pairwise choices among A, B, and C, people will often choose
non-transitively.  When they get the ternary choice, they have
to think more.

	It could be countered that the "true" preference is that evinced
in the ternary situation, but that line of argument can be continued
to demanding that a person's "true preference" is expressed only when
he knows all the alternatives available to him.  However, no-one ever knows
all the alternatives available to him, and systematic discussion
of all one's alternatives often leads to substantial changes of viewpoint.
Therefore, we had better not start down this slippery slope.