Explaining EAB with PCT

CSGnet Archive CSG1995
1

[From Rick Marken (950316.0910)]

Me:

Could you tell us what EAB stuff you do believe literally?

Bruce Abbott (950315.1855 EST) --

"Objection, yer honor: vague. Counsel is on a fishing expedition."

Why the paranoia? Do you think you're on trial? You sound like the defense in
the OJ case. Is there some evidence missing?

I've acknowledged your suggestion [that the logical relationship (red and
right) or (green and left) is controlled in the SDTEST study] on at least
two previous occasions and indicated that it is worth trying.

But your "acknowledgements" suggested to me that you thought that I was
making a theoretical suggestion. I was not. I was making a factual
suggestion. I was saying that some logical variable can be shown (using
The Test) to be a controlled variable in the SDTEST situation. No theory
is involved: just a demonstration that a variable (a logical one in this
case) is under control.

I am NOT proposing that the appearance of "stimulus control" in SDTEST can be
"accounted for" by a theory that says that the subject is controlling a
logical variable. I am saying that the phenomenon that is called "stimulus
control" and that appears to involve differential responding to stimuli is
really a different phenomenon -- it is a CONTROL phenomenon . What is
controlled in "stimulus control" is the perception of a logical variable. The
appearance of "differential responding to stimuli" (stimulus control) is just
a side effect of the disturbance resistance required to maintain that
perception in its reference state. Changing the disturbance (by, for example,
making the cursor move to the appropriate target all by itself on some
occasions) changes what the subject must do to control the perception -- and
we no longer see differential responding to the stimuli. The subject is
still controlling the same logical variable but we no longer see "stimulus
control".

I am trying give a clear, tangible example of why PCT is not an alternative
explanation of the phenomena of EAB (or of conventional psychology in
general). You have said that your goal is to show how PCT can provide a
better explanation of EAB phenomena than current models. The example of
"stimulus control" as a one of several possible appearances one might see as
a side effect of controlling a logical variable is meant to show that it is
important to figure out what phenomenon is to be explained before determining
what theory explains it better.

When I first got into PCT I, too, wanted to show how PCT explains all the
conventional findings in psychology. It took quite a while for me to realize
that this was a fool's errand. It is a fool's errand because PCT is about
control; conventional psychology is not. I hope the "stimulus control"
example above helps you understand what this means.

I wasted a couple years trying to figure out ways to account for conventional
results using PCT. I want to save you from having to replicate this waste of
time so that you can just get right into doing PCT research.

The message from PCT to EAB is NOT that PCT is a better model of their
phenomena. The message is that they have failed to notice the FACT that
organisms CONTROL. The first thing you should tell the EAB types is how to
test for controlled variables. Once they know that variables are under
control they can start trying to explain how this control is accomplished.
This is where they will find themselves doing what you are doing with the
data from SDTEST -- building a control model, the only kind that can control.

Best

Rick

2

[From Bruce Abbott (950316.1850 EST)]

Rick Marken (950316.0910)

Me:

Could you tell us what EAB stuff you do believe literally?

Bruce Abbott (950315.1855 EST) --

"Objection, yer honor: vague. Counsel is on a fishing expedition."

Why the paranoia? Do you think you're on trial? You sound like the defense in
the OJ case. Is there some evidence missing?

Well, Rick, it's like this.... (I'm trying to sound like Jerry Spence, the
folksy buckskin lawyer-consultant on Rivera Live--you have to read it
slowly) ... there's a lot of territory to cover there, and I'm not sure
which ideas from traditional learning and behavior I might find appealing if
they appeared to have the potential to contribute something to our modeling
efforts. I won't know until those situations arise--if any do--in which
some application might seem appealing. All I was trying to say is that I
want to keep an open mind about that possibility.

I am NOT proposing that the appearance of "stimulus control" in SDTEST can be
"accounted for" by a theory that says that the subject is controlling a
logical variable. I am saying that the phenomenon that is called "stimulus
control" and that appears to involve differential responding to stimuli is
really a different phenomenon -- it is a CONTROL phenomenon .

Rick, you're speaking doubletalk. The term "stimulus control" is simply a
name for an observed phenomenon--an empirical fact. The empirical fact has
been given a theoretical explanation by behaviorists in terms of
differential reinforcement. PCT offers another theoretical explanation, and
at present we're busy developing a set of PCT models to test that will
demonstrate how PCT offers a powerful alternative to the traditional view.
Whether you want to describe the end result as "accounting for" stimulus
control or "showing that it is really a different phenomenon" is just
semantics. The result is the same, however you wish to describe it.

When I first got into PCT I, too, wanted to show how PCT explains all the
conventional findings in psychology. It took quite a while for me to realize
that this was a fool's errand.

I appreciate your trying to save me from myself, but I seriously doubt that
I am on the "fool's errand" you describe. I don't know what "conventional
results" you tried to explain with PCT, but perhaps the fault lay in a poor
choice of results to model. The application of PCT to the phenomena of EAB
is direct and compelling. Rather than the dangers you seem to perceive, I
see an opportunity to make an enormous contribution to a field of inquiry
that has gone down the wrong path and is now finding that the trail is
getting thornier and more difficult with each step.

I wasted a couple years trying to figure out ways to account for conventional
results using PCT. I want to save you from having to replicate this waste of
time so that you can just get right into doing PCT research.

Thanks, but you have me confused. It seems to me that what we're doing IS
PCT research. In what way is it not?

Regards,

Bruce

3

[From Rick Marken (950316.2030)]

Bruce Abbott (950316.1850 EST) --

Rick, you're speaking doubletalk.

I can imagine that it sounds that way. When you listen to a Three Part
Invention for the first time it can seem like there are more than two
hands playing. When you listen to PCT for the first time it can seem
like it is spoken out of both sides of the mouth. Once you learn how a
Three Part Invention works, everything falls into place. The same
thing happens when you learn PCT;-)

The term "stimulus control" is simply a name for an observed
phenomenon--an empirical fact. The empirical fact has been given a
theoretical explanation by behaviorists in terms of differential
reinforcement. PCT offers another theoretical explanation, and at
present we're busy developing a set of PCT models to test that will
demonstrate how PCT offers a powerful alternative to the traditional
view.

Whether you want to describe the end result as "accounting for"
stimulus control or "showing that it is really a different
phenomenon" is just semantics. The result is the same, however you
wish to describe it.

I still think that there is an important difference between "accounting
for" a phenomenon like stimulus control and "showing that it is really
a different phenomenon". Let me try to explain my point of view one
more time.

I presume that you are familiar with the "rubber band" demo. One of
the fun things you can do with this demo is apply disturbances in such
a way that people trace out their own name. An observer who doesn't
see the controlled variable (the position of the knot) is likely to
conclude that people tend to write their own name when their finger is
in the end of a rubber band: the "name writing" phenomenon. Of
course, people are not really writing their name "on purpose"; the
name is just an accidental side effect (noticed by an observer) of control;
all people are doing is keeping the knot on target: controlling knot
position.

I believe that "stimulus control" has exactly the same status as the
name written while controlling knot position: it is a side effect of
controlling -- in this case, of controlling a logical variable. Just as
people no longer write their name when the disturbance to the knot no
longer demands this kind of action (in order to keep the knot under control),
organisms no longer display a correlation between responses and stimuli when
the disturbance to the logical variable no longer demands this kind of action
(in order to keep the logical variable under control).

I have suggested a way to change the disturbance to the logical variable
so that the correlation between responses and stimuli is no longer seen
-- even though the logical variable is still maintained in its reference
state. I will try to write the program to illustrate my point but I bet you
could write it in a flash. Just keep track of the state of the logical
variable over the course of a run; also keep track of the correlation
between stimulus color and mouse position during the run. Do this for
SDTEST as it is currently set up and then do it when SDTEST is set up so that
the cursor is moved to the appropriate target immediately after the cursor
changes color on a random 1/2 the changes What you will find is that the
logical variable is maintained (at "true") in both cases; but the correlation
between stimulus color and mouse position will be high in one case
(stimulus control) and very low in the other (no stimulus control).

Once you find that stimulus control is a side effect of controlling a
logical variable, I suppose it is possible to say that you have "accounted
for" stimulus control. After all, once you have discovered that name
writing is a side effect of controlling knot position, it is possible to say
that you have "accounted for" name writing. But you are not really
"accounting for" the phenomenon in theoretical terms; you are just
pointing to something observable that had been missed before. Indeed, you
are simply showing that a phenomenon (stimulus control, name writing)
doesn't really require a theoretical explanation because it is just a
side effect of a "real" phenomenon -- in this case, the phenomenon of
control.

Control is what PCT tries to "account for". It does this by proposing a
working model that replicates an observed example of controlling. Thus, we
have not yet "accounted for" the controlling observed in the SDTEST situation.
Accounting for this kind of control means designing a hierarchy (probably)
of control systems, the relevant perceptual functions, dynamics and
connections between these systems. The research you are doing will
provide much of the groundwork for developing this model. But the
model, once constructed, will not "account for" stimulus control; that
has already been "accounted for" -- it's an irrelevant side effect of
controlling a logical variable. What the mdeol will account for (if it
works) is control of the variables controlled in the STTEST task.

I don't know what "conventional results" you tried to explain with
PCT, but perhaps the fault lay in a poor choice of results to model.

I hope that I have explained why "poor choice of results to model"
is not really the problem. The problem is that there are no (few?)
conventional studies that are aimed at determining controlled
variables (like knot position or logical variables); these studies are
typically aimed at studying irrelevant side effects of controlling (name
writing or stimulus control) that happen to catch the attention of an
observer.

But I suppose you won't take my word for it; so I'll just wait around
and try to keep my mouth shut while you find out for yourself.

Best

Rick