<[Bill Leach 950710.18:37 U.S. Eastern Time Zone]
[From Bruce Abbott (950710.1225 EST)]
An impossible to deal with problem is that the invokation of "behaviour
is caused by present AND PAST reinforcers" one is faced with a totally
untestable assertion.
In its broadest sense, this says only the obvious, which is that
behavior depends not only on the current circumstances (feedback
functions, disturbances, etc.) but on past learning. This assertion is
no different than what PCT would hold. It is testable in the laboratory
where one has the ability to observe what has been learned and under
what circumstances.
I believe that PCTers also recognize that past experience "influences"
future behaviour though definately not in the fashion that seems to be
the belief of EABers.
So reinforcement theory is untestable because it appeals to previous
learning as a determinant of behavior but PCT escapes from this
conundrum because it appeals to the "influence" of past experience
through some other mechanims not stated? I don't follow...
No, no, NO, NOOOOO!!! And this just might be one of the most crucial
differences between the PCT and the EAB views.
PCT tests for and studies control of perception. When the researcher
has perceived a potential CCEV, the researcher will attempt to apply
disturbances and observe how the subject controls.
The disturbances should range in magnitude and "phase" (if possible).
Ideally, for studies in the realm that we are presently capable of
modeling, disturbances should not be such as to cause a loss of control
by the subject.
The PCTer is NOT particularly interested in HOW the specific control loop
set came into being. That such came about through "learning",
"experience" and doubtlessly contains random components in the process is
just plain accepted in PCT (for now).
The issue in PCT then, is to find examples of control, isolate "what" is
being controlled, study "how" the control is accomplished (as perceived
by the observer), postulate a control model that can emulate this
observed control and then TEST the model.
I presume that as experience is gained in the process some attempts will
be made to identify control processes that are "different" and are used
by either different subjects or even the same subject under different
conditions to control essentially similar CCEVs.
PCT accepts that there is little more that we can do currently to explain
why "goal" type references change except to assert that such changes are
due to past experience. Such neither "explains them" nor offers any
useful insight. While you may well be right that it is testable in the
laboratory, I doubt that sufficient knowledge of control exists to date
to allow meaningful testing.
Thus, PCT does not "escape" from the conundrum but rather studies _what_
and _how_ a subject controls and takes as a given that trying to
determine _why_ or change _what_ is either futile or will provide
misleading results.
Some comments on a message from Martin are included here because I think
that they are somewhat relevent to this discussion and Martin may not see
them anyway. I will be interested in Bill P.'s thoughts on this as well.
[Martin Taylor 950710 13:15]
... the organization of the control hierarchy. This may be all well and
good for one's own experience, but where does that leave you when
someone else claims to have a different experience that suggests a
different kind of organization?
I would suggest here that there really isn't any solid evidence to claim
that the organization of every individual of a species IS the same. It
just might very well be that there are different structures for the
hiearachy in different individuals.
Just the fact that an at least somewhat random process must exist in the
creation of the organization of the neural processes suggests that it is
quite probable that significant differences could exist.
Experience with control theory already tells us that many control systems
of vastly different organization can achieve exactly the same control
performance. If one then allows for some variance besides...
Even when all our perceptions are our own, including our perceptions of
what other people say, nevertheless if we believe in "science" we have
to have some kind of observations that "ought" to correspond to
observations other people will agree that they have made. We need to
work with perceptions of the environment, rather than with perceptions
of perceptions of the environment.
If we _need_ to do this then we are in serious trouble! Though I might
perceive something about the environment, when I am thinking about that
perception, the _thinking_ is indeed an entirely different perception and
related only by inference.
If I try to "experience" a controlled perception, what I am conscious of
may well be quite a different perception than the one that is actually a
perceptual input signal for some control loop (it may not be also, but I
doubt that at this point we really know one way or the other).
For "attending to" controlled perceptions I think that it is actually
even a lot worse than what I just stated above...
We probably NEVER control any perceptions in "isolation". That is, there
are always perceptions with references above any controlled perception
that we might be "attending to" and there are likely always other, not
necessarily related, controlled perceptions with "pre-emptive" priority
should their value deviate from acceptable limits.
Bill P.
What we're trying to model is the part behind the eyes, but the
phenomena we're trying to account for is in the part right here before
us.
Martin
And it is those phenomena that we have to try to select in such a way
that various people can agree to their existence.
In modeling terms this is presently not much of a problem as long as all
parties are in agreement that the phenomenon of interest is a control
phenomenon and that the application of the Test is the proper way to
study what is to be modeled.
When we do get to the point where different _Control_ phenomenon are
reported by different researchers we will have to deal with such
differences as any science should: investigate; attempt to understand
both what is different and how is that significant; what might explain
what is observed; and what are the implications to the theory?
In a very real sense, we have already encountered some of this. The
"performance" of different people in the oft run tracking experiments are
not only not the same between different people but differ some with
successive run with the same person in the observed and recorded results
much less those "actions" that are neither observed or recorded. As I
see it, the theory is NOT currently trying to explain this observed
variation in performance phenomenon (much less that which is not
explicitely observed and recorded) as long as the model can be made to
duplicate the "end point" performance to within an acceptable margin of
error.
At the moment the same model, differing only in slight parameter changes
and not in any fundamental way is used for all such modeling. At some
time in the future, it might well prove that it will be necessary to have
models that are structurally different for different people and/or
circumstances.
I think that we are currently severely restricted in that the PIF for all
of our models is significantly and substantially different from that of
our subjects. Some day...
-bill