Re: Misc comments on categorisation, problem solving, etc (Was Re: Aaron Somon's....)

From: Allan C Cybulskie (allan.c.cybulskie_at_yahoo.ca)
Date: 08/11/04 

Date: Wed, 11 Aug 2004 19:58:57 -0400

"Wolf Kirchmeir" <wwolfkir@sympatico.ca> wrote in message
news:tRfRc.50631$Vm1.1358755@news20.bellglobal.com...
> Allan C Cybulskie wrote:
> > "Wolf Kirchmeir" <wwolfkir@sympatico.ca> wrote in message
> [...]
>
>
> > The more and more I read about this whole behaviourism debate, the more
I
> > realize that the key to this entire picture for intelligent behaviour is
the
> > goals and desires of the agent, and the more I realize that you and
Glen --
> > perhaps unintentionally -- seem to minimize that role which leads your
views
> > to not ring totally true.
>
> And what's your evidence that an "agent" has "goals and desires?"

Well, it was derived from the entire rest of the post that you cut from
before this, but are you really denying that you have at least functional
structures that we commonly refer to as "goals and desires"?

>
> > Intelligent beings set, for the most part, their own rewards and
> > punishments. There are few, if any, biological or social constraints
that
> > cannot be turned by the views of the agent against what they're
intention
> > was. Pain can be considered a reward for some people. Hunger can be a
> > reward for people who are on a diet, and encourage the behaviour even
though
> > it's supposed to discourage it. Pleasure can be a negative
reinforcement
> > for someone trying to avoid it. And so on, and so forth.
>
> I don't understand wht you think you are refuting. That the same
> experience/sensation/??? can be both a positive and a negative
> reinforcer, depending on the situation, is just what one would expect -
> if one understood that the situation is a controlling factor in behaviour.

Well, it isn't the "situation" that I was referring to, unless your meaning
of "situation" is far different from everyone else's. To talk about
"situation" would normally only refer to what was going on externally, and
not to the internal goals and desires and mindset of the individual, which
is what I was commenting on as determining the reinforcer. For example, the
audience could applaud politely or loudly or snap their fingers or cheer or
whatever, and it might not impact the reinforcement at all. But the agent
deciding that they weren't aiming for applause in taking that action will
change the reinforcement of it greatly. So what do you mean when you talk
about situation? Everything?

>
> > In order to interpret the behaviour of the agent, and affect it, you
have to
> > understand their goals and desires, or else they'll turn your attempts
at
> > conditioning completely around on you.
>
> This apears to a roudnabout way of saying, "In order to change the
> agent's behaviour, you have to understand what environmental factors
> ("contingencies", in Glen's language) affect that behaviour.

Only if you consider everything -- including mental state -- as
"environment", which is non-standard and also leaves you in a rather poor
position, since if you define all possible factors as "environment" and all
possible actions as "behaviour", you leave yourself saying something that
we'd all agree with -- that that sort of environment is all that affects
behaviour -- but that we will quickly reject as too general a language to
make any progress in actually understanding what's going on. For example,
it would make much sense at this point to differentiate between the external
and internal causes and classify based on that distinction -- which would
bring us back to the classical description of things.

> We can know
> an agent's "goals and desires" only from its behaviour, which, in the
> case of a human, may include an assertion that it has certain goals and
> desires.

And how can we know the goals from the behaviour? It HAS to be because the
behaviour is critically determined by the goals themselves; in short, the
behaviour occurs BECAUSE OF the goals and desires. This makes goals and
desires critical parts of the behaviour themselves, and thus ultimately what
is of most interest in attempting to understand and control behaviour.

> Given that humans are notoriously skilled at deluding
> themselves or lying about themselves, such assertions aren't of much
> value without other evidence.

But we are even more notoriously skilled at deluding OTHERS, so relying on
what we show in terms of behaviour is likely much worse. We are more likely
to know what our own goals are than some other observer is. Some times,
yes, we can delude ourselves, and behaviour other than self-reporting can
help us see those problems. But that does not change the fact that, for the
most part, the self-reporting is less likely to be delusional than anything
else.

This is one of my objections to David's attempted work with prisoners, since
it relied on observing their behaviour and working with rewards for "right
behaviour" while ignoring the idea that intelligent prisoners will be able
to figure out what behaviour you want, act that way to get the rewards,
while never actually changing their behaviours at all -- which is what the
plan wanted to achieve.

> > In short, an agent will only react
> > to the rewards and punishments offered IF THEY WANT TO, at least for
> > behaviours that should be intelligent.
>
> This sounds like a tautology.

So what? That is not any sort of objection. Intelligent behaviour does
indeed rely on the idea that the beliefs and desires of the agent are
rationally considered and reasoned out to an optimal action. You can show
that such behaviour isn't really there, but if I say that that's what
intelligent behaviour means and so ask that candidates for such behaviour
have those qualities, it does no good to argue that I'm asking for things
according to the definition of intelligent.

>
> > Intelligent conditioning is
> > conditioning in sync with the desires of the agent; brainwashing is
attempts
> > to condition without that synchronicity. And simple behaviourist
techniques
> > don't necessarily promote a sync between desires and outcomes.
>
> Well, then , don't use simple techniques. There are plenty of more
> complex ones. In any case, if (for example)I want to increase sales in
> my store, I don't want to "condition" shoppers to spend more money, I
> want to trigger the pre-existing shopping behaviours that result in
> spending more money. IOW, theres a lot more to successful behaviour
> shaping than mere conditioning.

But I fail to see why what you say is any different than: "We know people
tend to believe X, and so if they do we can do this in order to play to that
belief (or desire, or whatever)". And since this part of the post was about
conditioning, I still say that most behaviourist conditioning does not
necessarily promote that sync.

Relevant Pages

  • Re: Misc comments on categorisation, problem solving, etc (Was Re: Aaron Somons....)
    ... that we call "goals and desires". ... >> is what I was commenting on as determining the reinforcer. ... You are demanding evidence to show that people are good at lying to ...
    (sci.cognitive)
  • Re: The structure of a self-conscious mind
    ... > about conditioning techniques and their use in controlling emotional ... you give to them are useful or correct (like goals, ... And I believe if you duplicate that type of reaction in a computer, ... >>> my personal experiences of mind. ...
    (comp.ai.philosophy)
  • Re: Misc comments on categorisation, problem solving, etc (Was Re: Aaron Somons....)
    ... WK:And what's your evidence that an "agent" has "goals and desires?" ... ACC: Well, it was derived from the entire rest of the post that you cut from ...
    (sci.cognitive)
  • Re: The structure of a self-conscious mind
    ... > because it seems to capture our own experiences and makes internal events ... :) Ah, the old mind problem. ... >> with operant conditioning to think that is better. ... the reason we talk about goals is not because ...
    (comp.ai.philosophy)
  • Re: Temporal Learning
    ... >>bot to more rewards. ... > use that goal as part of a reinforcement strategy for learning how to ... Newly created goals should be given a positive value, ... you give it a critic which rewards specific results. ...
    (comp.ai.philosophy)
Quantcast