Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!udel!gatech!howland.reston.ans.net!news.sprintlink.net!sundog.tiac.net!wizard.pn.com!Germany.EU.net!EU.net!uunet!psinntp!scylla!daryl
From: daryl@oracorp.com (Daryl McCullough)
Subject: Re: Folk psychology (was: Is Common Sense Explicit or Implicit?)
Message-ID: <1994Oct1.033151.26051@oracorp.com>
Organization: Odyssey Research Associates, Inc.
Date: Sat, 1 Oct 1994 03:31:51 GMT
Lines: 183

rickert@cs.niu.edu (Neil Rickert) writes:

>>Take almost any example of a simple interaction involving two humans.
>>If you tell somebody that "You left your car lights on." or "You left
>>your car window down, and it's starting to rain." and you can predict
>>with good accuracy how that person will react. Why is that? I think it
>>is because you assume that the person has certain desires (not to run
>>down his battery, not to get the inside of his car wet), and that he
>>will act on them based on the information he has available. If
>>somebody asks you "Why did that person go back to his car?" you will
>>be able to tell them "To turn off his lights" or "To roll up his
>>windows".
>
>It boggles the mind!
>
>As you said, "you assume that the person has certain desires .."
>Quite clearly, the prediction and behavior is correlated with your
>assumptions about the person's desires.  You have no evidence that it
>is correlated with the person's desires.

All I claimed was the predictive success of reasoning in terms of
desires, beliefs, etc. And you seem to be unable to dispute that.
As your remark "It boggles the mind!" shows, you are unable to see
the success of such reasoning because it is so commonplace that it
seems trivial.


>The actual desires that the person might have played no role
>whatsoever. Your prediction was based entirely on your assumptions
>about desires, rather than the desires themselves.

Right. Predictions are *always* based on a model of reality, and not
on reality itself, which we can never get access to directly.

>Thus folk psychology is not involved in the prediction.

Yes it is. Folk psychology is the model that I am using for human
behavior, and the prediction comes from examining that model.

>The sort of explanation you describe is about the same as "The clouds
>were dark and angry, and it rained because of their anger."

What's the point of these rock and cloud analogies? If cloud behavior
were as rich as human behavior, and if the cloud showed signs of
remembering past events, of planning for the future, of having desires
and motives---if I could have a conversation with a cloud to *discuss*
its anger and its reasons for it, possibly talk it into seeing some
counselling for more constructive ways to show its anger---if a cloud
behaved humanly, in other words---*then* I would probably use folk
psychology to understand cloud behavior, as well. But since it doesn't,
I don't.

>We make predictions about rain because we have a great deal of
>statistical evidence from similar looking clouds that we have seen in
>the past.

Right.

>We make prediction about human behavior, because we have a
>great deal of statistical evidence about similar human behavior we
>have observed in the past.

Wrong. In the case of clouds, we can get good statistical data because
there is a very small number of parameters: the temperature, the wind
speed and direction, the color and size of the clouds, etc. In the
case of human beings, the number of parameters is enormous. There is
no way in our short life spans that we can come close to getting enough
statistical data to even start making statistical predictions. I don't
actually think that informal human reasoning ever involves statistics.
What it involves is model-building, and reasoning within models. And
our model of human beings (or at least mine) involves mental attributes
such as beliefs, purposes, etc.

>According to the folk psychology theory, here is the apparent
>situation:
>
>	In our experience, we have had the opportunity to gather
>	statistical evidence about human behavior.  However, we have
>	no way of knowing the desires of another person.  We make
>	predictions about behavior.  It is said that our predictions
>	are based on the desires, which we could not possibly know,
>	and it is said that the statistical evidence we have had
>	every opportunity to gather is irrelevant to the issue.

Sigh. I see that the heart of our disagreement has nothing to do
with folk psychology. We don't agree about what empirical science
is all about. To me, science is about building models to explain
the universe and make predictions about it. A theory is judged to
be successful or not in terms of these goals.

Statistical data by itself is not a theory. Statistical data by itself
can make no predictions, because the exact situation never comes up
more than once. (Not if the situation involves things as complex as
humans, as opposed to dice.) There has to be an element of theory in
statistical predictions, at least to the extent of classifying which
situations are "similar" to which other ones.

>If this counts as empirical science, then astrology and creationism
>should surely also count.

If a statement like that counts as an argument, then so should this
one: I'm right, you're wrong, so there! Sorry, I guess I'm getting
tired.

>>          These are trivial predictions, but if you consider how very
>>complex human behavior is, it is a tremendous success to be able to
>>make such predictions.
>
>I am not disagreeing with the ability to make predictions.  I am
>merely pointing out that folk psychology has nothing to do with it.

Well, you are wrong. Folk psychology was what I used to make the
prediction. Give me an alternative explanation that doesn't involve
beliefs or desires.

>>                               If you had answered the question "Why
>>did that person go back to his car?" with an answer such as "Because
>>synapses fired, causing an electrochemical pulse to travel to his leg
>>muscles" people would have considered you insane. Not that the answer
>>is wrong, only that it is inappropriate, at the wrong level.
>
>I agree that, in most cases, this would be an inappropriate
>explanation. But it does not follow that the intentional explanation
>is correct.

Then give me an example of an explanation you would consider correct
and appropriate.


>>To me, folk psychology defines the subject matter of cognitive science
>>and Artificial Intelligence.
>
>If this is so, then cognitive science and artificial intelligence are
>just as valid as astrology and creationism -- no more, and no less.

Sigh. Since I have no idea what you mean by "valid", and since I am
getting tired of your comparisons, I won't even argue with you. But if
you want, I can just speak for myself, rather than entire fields: what
*I* want out of Artificial Intelligence is a machine which behaves as
if it had goals, beliefs, plans, etc. Why *I* am interested in
cognitive science is because I want to know how purposeful, reasoning
behavior arises from nonliving, unintelligent matter.

>>                             A system will never be considered
>>intelligent or conscious unless some kind of folk psychology applies
>>to it.
>
>That may be correct.  But that is only a backwards way of saying that
>intentional talk about a system will not be socially acceptable until
>there is a social consensus that the system is intelligent.

No, I said it the way that I meant it. Once again, I will personalize
this: *I* will never consider something intelligent until I can understand
its behavior in terms of purposes, beliefs, etc. That is part of what I
mean by intelligence.

I don't understand why you don't object to the word "intelligence".
That is just a folk-psychology term, after all.

>If social consensus is to be the standard for judging AI, then all of
>the anti-ai arguments of Searle, Dreyfus and others are quite correct,
>and we might just as well give up right now.

That's a bunch of non-sequiters (if I spelled that right). I didn't
say that it should be based on social consensus, I said it was based
on the extent to which folk psychology applied. There doesn't have to
be any consensus; each person can decide for themselves. If you are
looking for an objective standard that AI has been achieved, then you
are going to be disappointed. There is no such standard, and there
never will be, (in my opinion).

Daryl McCullough
ORA Corp.
Ithaca, NY









