DEMO1: RESULTS OF EXERCISE - SOCIOLOGY 354 F'94
UNIVERSITY OF SOUTH CAROLINA - PROFESSOR CHARLES W. TUCKER
. . .
Data sets omitted here.
Chuck is doing something important for PCT. He is asking lots of people to
sit down for a little while and do come controlling, in a setting where
moment-by-moment values of some important variables are saved and analyzed
in a way that makes them useful for control scientists. Aside from a few
fanatical PCT modelers, no one else has done that on a big scale. Chuck is
getting a big dose of empirical evidence about the phenomenon of control
and the results are worth his effort.
He has seen that:
It seems that few of these students had problems with
control a line on on the scree whether it was with CT or
PT, several had difficulties with "relative size" (9935,
0244), several with "orientation" (0457, 9935) one with
"shape" (7654) and one with "numbers" (9935) but all but
a few (7654, 7315, 8962) had difficulty with "pitch of
sound."
Chuck is seeing that, contrary to the expert opinions of our many editors and
reviewers, all people do not perform equally, nor do they all perform
well, nor do they each control all variables equally well. Further, even
"poor" control is obviously different from no control at all. Those are all
important facts to keep in mind during any discussions about the phenomenon
of control.
The goal of PCT-science is not to find instances of perfect control. We are
much more excited about data like Chuck's, that show differences between
individuals and between control of different variables, than about someone
who performs nearly perfectly. Data like Chuck's let us show how even the
simplest PCT model can match and predict the performance of people who
differ widely in their proficiency at control.
I would (and I suggest all who do this) have to
observe these students while they are doing these exercises
to get some idea of what their difficulties might be with
these tasks.
This brings tears of joy to the eyes of those who rage against the darkness
brought upon the land by proponents of group statistics and average
curves! Your students are individuals. If you want to know what each one
does, you must watch them all, one at a time. You cannot use a summary of
group data, no matter how sophisticated, to learn anything about any one of
those people. But, as Phil Runkel explained lucidly, when we carefully
observe one person as our "specimen" of the species, and when we we develop
a good working model for that one person, we stand a chance of explaining
the performance of all members of the species. Averaging across the
species tells us nothing about how any individual functions; understanding
one representative member of the species tells us something about how each
member of the species functions.
An aside for Bruce Abbott. Observing and explaining the performance of many
ordinary (to use a politically incorrect word, "normal") individuals like
those in Chuck's study tells us a lot more about control than if we
deliberately start with individuals whose performance is compromised in one
way or another. For decades, psychologists have asserted that we can learn
more about how to explain behavior if we begin by studying cases of
psychopathy than by studying unimpaired individuals. That's a lot of
poppycock. If you don't even know what a normally functioning VCR does,
you will learn nothing about VCRs by studying the performance of those that
were tossed off the assembly line because of major defects. And VCRs are
simple things compared to living systems.
But I have thought about the "pitch of sound"
task and remembered that it is the onel one of these tasks
which is not visual but where the GOAL CHANGES EACH TIME.
Not exactly. The "reference" tone is the same frequency at the start of
every run. But you are right in another very important sense. The person's
memory of the tone probably does vary throughout each run, so each person's
reference perception changes across time, probably in different ways for
different people, with an average of "no change" across many people and
many runs - yet another reason to reject averages and focus instead on each
person.
In all other tasks the purpose is to do the "same" thing
each time and the task is there before one but in the
"pitch of sound" task the sound CHANGES each time so if
one could "remember" the sound on the last "run" it would
be of little help on the next "run." This may not only be
a "hearing" problems but a "remembering" one also. It is
also difficulty to get the idea of changing a sound by
moving a mouse while changing an "image" on the screen with
a mouse is more familiar to anyone who has used a mouse.
Thus we should be dealing with different types of "images" as
well as their different forms to see what a person might be
able to control.
Yes! Isn't it fun to look an some real data -- some empirical evidence
about the phenomenon of control? All sorts of irrelevant speculations fall
away and you begin to notice the important facts that must be explained.
You are immediately in the world of "what is," rather than the world of
"what I believe must be."
You are on the same trail I tried to examine briefly in my paper on mimicry
and repetition, in _Closed Loop_. handle, with movements
of the target on the screen. (The person's hand is blocked from view; the
person matches felt movement and position against seen movement of another
variable.) The correlation between handle and target positions is still
relatively high and positive, but always smaller than it is for seen cursor
and seen target. And the rms error is bigger than for seen cursor and
target.
Next, after the person has extensive practice trying to match felt movements
to seen movements, the screen is blanked completely and the person tries to
reproduce the felt movements that occurred in a one-minute run during
practice. The person tries to repeat the pattern of hand and handle
movements, from memory. Averaged cross a few runs like this, the
correlation is "zero" between positions of the person's handle and the
actual positions that would reproduce the original movements. Of course,
the rms error is immense. People have a devil of a time making their
movemenmts repeat, exactly, even for runs as short as one minute. Like you,
Chuck, I suspect the "problem" is in the variability of their memory for
exactly what they want to do, at each moment in time.
Your results are getting at some of the same phenomena as mine, Chuck. We
seem to have agreement in our observations that there are replicable
differences in the *environmental* results when people (a) control the
relationship between two variables simultaneously present in the same
sensory modality, (b) control the relationship between two variables
simultaneously present in two different sensory modalities, and (c) control
the relationship between two variables, one of which is present in one
sensory modality, the other in memory. All good grist for modeling.
Another thing occurred to me *after* I put my results in Closed Loop. My
results, like yours, describe what an *observer* sees when a person controls
under the three kinds of conditions I described above. From the observer's
point of view, the person controlled well when two seen variables were
present, controlled less well when felt position of one variable was matched
against seen position of another, and controlled least well when felt
position of one variable was matched remembered position of the same
variable. But that is all misleading. From the *person's* perspective,
control was fine all the time. Under condition (a), the cursor looked like
it was in just about the right place all the time; under conditions (b) and
(c), the hand felt like it was in just about the right place all the time.
Each person controlled his or her perceptions "just fine." External
observers saw things differently. The controller does not sense momentary
drift and change in the setting of the reference perception; those changes
become part of the net disturbance to the controlled perception and the
person simply keeps on controlling -- everything is fine. Meanwhile, the
external observer can be going into fits and raving about how terribly
people perform under conditions (b) and (c). This bias in favor of the
observer and against the person is like the one that permeates traditional
behavioral research, where the literature is filled with data that
purportedly reveal how poorly subjects perform and how their results are
filled with "error." True to form, PCT is on the side of the insider. 
I would suggest the everyone try (if just with yourself)
to do these tasks and see what the results are for you and
others.
Yes! Yes! Yes! *Then* come back and speculate to your heart's content,
your speculation disciplined by the magical elixir of empirical observation!
While putting this together for this mailing I thought of
the publishing of these results( Tom challenged me to
"join the club" of "rejectees" even though he doesn't know
that I am a charter member of such a "club" in my own
discipline!)
I know you are, Chuck. I was inviting you to "come out," and reap the
rewards -- at the next CSG meeting, we can all (all four or five of us) sit
around, sip some cider, and exchange rejection myths.
and realized that the purpose of these exercises
is simple to devise ways to find out if a person can control
some variable that you asked him/her to control.
Yes! Look first, then talk.
If the
person is unsuccessful it means that they just can not "set"
that variable for themselves, that they can "set" it but don't
"control" it very well or that there are other disturbances
that disrupt with the "controlling" of the person. This is
just preliminary to finding out if a person behave as the
living control system but only would have serious implications
for the model if no one could ever "control" anything. That
hasn't happened yet.
Ah, Chuck. I want to give you a big hug!
By why would anyone publish such
"technical" stuff? Maybe we should start our own journal
called THE JOURNAL OF LIVING CONTROL SYSTEMS published by
The Society for the Study of Living Control Systems. Naah.
Naah!
Later,
Tom
···
In Message Mon, 9 Jan 1995 15:01:29 EST, "CHARLES W. TUCKER" <N050024%UNIVSCVM.BITNET@uga.cc.uga.edu> writes: