[Martin Taylor 951128 14:00]
Bill Powers (951128.0805 MST)
I didn't really suppose you to mean what you said, before, but it
could certainly have been misleading to quite a few readers, and it
definitely took _me_ quite by surprise, despite my years of
interaction with you. I'm not going to requote and comment,
because the above statement is succinct and all that is necessary.The sentence you don't want to requote is
That is the main thesis of PCT: the ONLY reason for behavior is to
control perception.
No it isn't. It is the last sentence of
In the case of the rat and the shock, there is no feedback effect fast
enough to counteract a single shock. But the "proximal stimulus" in this
case is not the effect of a single shock. It is an effect averaged over
time. The running can vary in such a way as to counteract (to some
extent) this average effect, so it is the average effect that is part of
the loop. The only reason there is running is that it does have an
influence on this average effect; if it had no influence at all -- if
running or not running made no difference in what the rat experienced --
there would be no running.
which immediately predeces the one you requote. In the message on which
you comment, I shifted the emphasis in:
That is the main thesis of PCT: the ONLY reason for behavior is to
control perception.
from ONLY to REASON, which may be why you took it to be the one I didn't
want to requote.
I didn't want to requote "if running or not running made no difference in
hat the rat experienced, there would be no running," because you seemed to
have corrected it, and I thought that to requote it might mislead some
readers.
From this sentence, you seem to have manufactured "All perceptions are
controlled by behavior." What took you by surprise was your own hasty
misreading of the sentence.
No. What I (mis)manufactured was "all actions in fact serve to control
perception," a denial of the existence of the very side-effects about which
we have been having a somewhat involved discussion.
But since you brought it up, it's probably an important element in
why the cat wants to get out of the box. If there were no error
involved in its being in the box, as would probably be the case if
it wandered in of its own accord, there would be no reduction of
error when the door opened and nothing to learn. So, yes,
overwhelming force may be a component of the whole experiment, but
it was not what I was referring to in what Bill L was commenting
on, nor does it apply when the cat is actually in the box."Nor does it apply when actually in the box???" It is the enormous error
induced by forcibly placing the cat in the box against its will that
motivates all that follows, isn't it?
Isn't that precisely the point I was trying to make? That it is not the
"overwhelming force" but the error that is important? I'm glad you agree,
even if you do cast your agreement in the form of a critical comment on
what I said.
We have to guard against tunnel vision when
looking at experimental results. What happens before and between
experiments may be as germane as what happens during them.
Valid comment. To be kept in mind.
···
-------------------
I'd like to continue our mutual exploration of the many implications of
side-effect components of outer control loops. Rick wants to open up
discussions of larger social issues, and side-effects may provide one such
opening, among other possibilities.
----------------
Bill Leach 951127.18:48
However, as you also mentioned (I think), that while "learning" might
have taken place, the learned control may have nothing more to do with
moving the stick than "pure chance". This is testable of course ...
One of the issues in the discussion is whether, or rather, under what
circumstances it is testable. I don't think the interchange with Bill P.
has yet resolved that question.
I'm arguing that the experimenter can never see whether the cat learns
about the stick as a mechanism for getting out of the box (forget the
PCT aspects of the problem--this is more general).Not by just repeating the experiment but then I believe that has been
Bill P.'s primary point in the whole discussion.
And one of mine, which I have tried to illustrate in a variety of different
ways. There's no difference of opinion about that, with the caveat that
as Bill P. says, apparent agreement may hide a quagmire of different
underlying assumptions about what is meant by different statements and
words.
My objective in this discussion is to consider the implications of the
FACT (if one assumes the essential correctness of HPCT) that there will
exist control loops in which some link involves a side-effect of control
of some lower-order perception. The higher (outer) control loop sets
a reference for some lower order perception, but it is the side-effect
of that control under constant environmental conditions that actually
affects the perception controlled in the outer loop. (Such control loops
must be very fragile when confronted by changes in the feedback structure
of the environment, but as in the hypothesized illustration of the cat,
equivalent control loops that still are based on side-effect may be rebuilt
during reorganization.
This fact may be hard to detect either by the actor or by an outside
observer/experimenter, inasmuch as the existence and magnitude of the
side effect is highly correlated with the existence and magnitude of
the output that influences the lower-order perception. Remeber that the
actor does not perceive the action. The correlation holds so long as the
environment maintains the same structure of available feedback paths,
and it leads to ritualized behaviour that usually works, so long as the
environment doesn't change too much. The ritual doesn't always work, but
when it doesn't there are always "reasons" -- extraneous influences -- to
explain away the failure, whether the actor or the IV-DV experimenter is the
one doing the explaining.
It is that kind of implication that I'm trying to explore--there seem to
be many such, of which the discussion has touched on a few.
I guess I am
"still hung up" on the idea that when running an experiment one is quite
literally "pitting one control system against another" and recognizing
that fact is absolutely essential to reaching a valid conclusion
concerning the results.
This comment intrigues and puzzles me. Could you elaborate? I would have
thought that at least with human subjects in an ethical experiment, the
subjects are willingly doing what they think an experimenter wants them
to do. I don't see in what respect running such an experiment necessarily
depends on pitting one control system against another. Even in the case of
the cat unwillingly placed in the box, the conflict occurs outside the bounds
of what is formally called the experiment (but see Bill P's "valid point"
above). In the box, the cat is (to anthropomorphize) solving a problem,
and I'm sure we can think of myriads of problem-solving situations into
which we get ourselves without having been overtly subjected to overwhelming
physical force. My claim is that it is the error involved in the situation
(cat not perceiving what it wants to perceive) that matters, not how that
error came to exist. Conflict between control systems is one reasonably
sure way to induce error, but it's not the only way.
Martin