Re: Challenge to the behaviourists #2

From: Allan C Cybulskie (allan.c.cybulskie_at_yahoo.ca)
Date: 09/19/04 

Date: Sun, 19 Sep 2004 08:31:45 -0400

"Neil W Rickert" <rickert+nn@cs.niu.edu> wrote in message
news:ci4bl1$ack$1@usenet.cso.niu.edu...
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> >> >Um, since we generally want to get artificial intelligence to have
human
> >> >intelligence levels, we'd expect to be able to predict and control
their
> >> >behaviour just like we would for humans, and that ability would help
us
> >> >determine if any AI project was actually successful. Therefore, it is
> >not
> >> >off topic at all.
>
> >> However, we do not control and predict the behavior of humans. It is
> >> a characteristic of intelligent behavior that it is not predictable,
> >> and that it resists control.
>
> >I'm not sure of that. You seem to be equating "free" and "intelligent"
> >here, which might lead you to think that it couldn't be predicted or
> >controlled.
>
> Why do people play chess? If we could predict human behavior, we
> would know the outcome before the game begins.

We can, to an extent, if we know the how each person plays and what
strategies they tend to employ. The only novel thing is when we get
creative to overcome the pre-determined games.

>
> Why do we hold elections? If we could predict human behavior, we
> would know the election outcome before the vote.

And we generally are indeed able to predict these things from simply polling
a small number of people and extrapolating to the whole ...

>
> Why do we hold trials (legal proceedings)? If we could predict human
> behavior, we would know the verdict before the trial started.

Well, this would depend on what evidence was actually presented, and how it
is presented, which cannot be determined until after the presentation of
evidence. After such a presentation, we are indeed quite good at predicting
verdicts. And most lawyers rely on being able to predict which cases will
or will not win -- to a reasonable accuracy, although not perfectly -- for
their livelihood.

>
> > As an example
(adapted
> >from a philosophy of science text), it might be the case that I don't eat
> >Sloppy Joes because they make me slightly sick, but do eat poutine
because I
> >like it. This would be an intelligent behaviour. Now, knowing this, you
> >could predict that if the cafeteria was serving poutine and Sloppy Joes
that
> >I would choose the poutine.
>
> No I couldn't. You might decide to skip lunch that day. Or you
> might decide to eat elsewhere. Or maybe you would choose the sloppy
> joes to test whether you had outgrown that reaction.
>
> The world is a far more complex place than what is depicted in the
> "Just So" stories that epistemologists tell.

Please don't pull a "David" here. I stated quite clearly that you could not
get absolute predictions, but that you could get pretty good ones. Yes,
sometimes they will be in error, but for the most part it would be either
things that we didn't know about or irrational behaviour. Your attack here
depends entirely on my claiming you could get absolutely correct, guaranteed
predictions, which I explicitly said I was NOT claiming.

Intelligent behaviour is easier to predict than, say, emotional behaviour
since it generally is based on proper facts and reasonable desires.