Tom Bourbon [941017.1014]
FROM CHUCK TUCKER 941017.0912
Let me start with a bold assertion: compared with any area of
"research" in the so-called "behavioral science" the "research"
done under the name of PCT is quite meager and unconvincing.
Chuck, thanks for picking up on the topic of PCT research. To help me
follow your thoughts on the subject, could you tell me what you _do_ find
convincing in traditional research? I need to know that, so I can better
appreciate what you identify as problems in research PCT. What makes
traditional research convincing to you? Do you see anything of value in the
research done so far on PCT? If so, what?
If you think the amount of quantitative research on PCT is meager, then we
agree. I repeat my invitation to everyone who shares that idea: start doing
some of the work, rather than sitting on the sidelines and carping at the
few people who have taken PCT seriously enough to dedicate their time and
resources to doing quantitative research. Chuck is doing some of the
groundwork that will prepare him to join in the research effort, and he is
identifying some good topics to study. Where are all of the others who
should be doing the same thing as Chuck?
All of it has been done under the circumstances where the
experimentor has defined the reference signal, the disturbance,
the mode of behavior to be used to counter the disturbance and
and the meaning of the empirical results.
I don't quite agree with _everything_ you said here. For discussions about,
and examples of, people adopting various reference perceptions during a
simple tracking task, see "Models and their worlds," and "Mimicry,
repetition and perceptual control," both of which are in _Closed Loop_, and
"Program-level control of a sequence of perceived relationships," which is
my paper in the proceedings of the meeting in Wales. Also, look at Bill
Powers's paper in Psychological Review in 1978, and his chapter in Wayne's
book. And don't forget Rick's "Mind readings." These examples all include
variations on simple control tasks, but they demonstrate a deeper analysis
of control behavior than is implied in your remark.
All of that aside, I _agree_ that this database is far too thin, given that
PCT has been on the scene for a few decades. More people should get off of
their duffs and go to work.
In none of our
studies using the "tracking design" as a "model" either as
research or as demonstration do we use THE TEST to "find" a
"controlled variable" that we (the experimentor, research or
teacher) DON'T ALREADY KNOW. This can be said for all of the
the computer programs devised by Bill, Rick, Tom (I have used
all of them in research except Tom's)
You have not tried the tasks I cited above. In them, people can adopt
different reference perceptions. That is also true for the two-person
programs I sent to you. When you run them, suggest to the participants
that they decide for themselves which relationship to create and maintain
throughout the next two runs, then watch the results. Within the limits I
was able to achieve with my programming back then, you will discover that
after the first run the program estimates their reference perceptions and
their integration factors and uses those parameters in the PCT model to
predict the results of the second run, when either the target or the
disturbance or both are different from the ones in the first run. Chuck,
this _is_ The Test. If the people reproduce their _results_ from the first
run, but with different movements, and if their performance on the second
run is like that predicted by the model, then The Test has confirmed the
possibility that we know the controlled relationship. (This theme is found
in _many_ of our tracking studies: Task 1; estimation of parameters;
prediction of Task 2 (when there are unpredictable changes from the
conditions in Task 1); Task 2; compare predicted and actual results from
Task 2. This procedure _is_ The Test.)
and all of the demos
that I and all of us (see issue of CL for example) have use
to "show how a negative feedback control system a la human
being" works.
Yes. That is one of our intentions, and we always urge others to take up
part of the work and build on our quantitatively solid demonstrations.
We have tried to provide a firm foundation on which others could build.
Where are the other builders? (By the way, PCT _is_ a theory intended to
explain the _phenomenon_ of control. We _must_ first demonstrate clear
examples of the phenomenon, then show that the theory can explain and
predict the phenomenon. Can anyone else suggest another way for us to start
building an empirical base for PCT?)
The Crowd program is one that shows how "simulated
actors" as "control systems" engage in "individual locomotion"
that appears to look like what Clark and I have have observed
and that Clark and Ron did in their research (a publication
by the way which is not often noted in the PCT literature).
But, as I noted in a post on 0929, there are serious problems
with the program when it comes to all but thes simplist movements.
The program does not YET simulate most of the collective locomotion
forms that we have observed.
Great! You have seen the model, you have seen the phenomena the model was
intended to explain, you have seen where the model succeeds, and you have
seen where more work must be done both in the field and in the modeling.
That is _exactly_ the kind of thing we hope will happen -- _if_ you follow
through and do the additional work. I believe you will. 
THE POINT: Until we can do research on "reference signals" that we
do not strongly influence or that we do not simply use as a
"demonstration" very few people who might be interested in making
sense out of what people do to, with, for, and against one another
will take us seriously.
The "social tracking" studies my students and I presented at CSG
meetings over the past eight years or so were intended to demonstrate (a
good word, in my lexicon) at least a few ways others might adapt PCT
modeling to their interests, or adapt their interests to PCT. I am stil
waiting to see them do so.
We can continue to say "we don't see that
as a behavior" because you have not characterized it in PCT terms
but that does not work or it is not convincing (that is what is
happening on the net recently in the conversations between Abbott
and others).
Chuck, I believe the point is that the data in most traditional research are
not in a form that allows us to know which perceptions the subject's might
have been controlling. That is a fact, not a matter of whim or bias on our
part. If other people do not study control _by their subjects,_ or worse if
they design their research so that any evidence of control by subjects is
obliterated, then the data cannot be used as clear evidence for control and
we cannot apply the PCT model. If there are no data that allow us to see
what the subjects were controlling, then we cannot see what they were
controlling. That's all.
Every time someone suggests that an experiment be
devised that does use "tracking" as research design there is
resistence.
I don't know which posts you have in mind. Can you cite some of them? I
don't remember seeing resistance on this point.
How about some advise on the design for an experiment
which simple involves people interacting with each other?
Look back at some of the things I suggested after Clark cited Sherif's
research. I tried to identify several ways that PCT-ish social scientists
could follow up on the hints of controlled perceptions that we find in
Sherif's work. Also, think back over all of those perssentations and
demonstrations of "social tracking" that we gave at CSG over the years; we
always talked about how our simple tasks could be turned into larger "more
meaningful" projects. Go to it!
Later,
Tom