Thanks, Tom; Epistemology

[From Rick Marken (930512.1000)]

Tom Bourbon (930511.1712)--

Ken, in those posts you alleged that several of us on the net were
attacking you and were denying the importance of things social. I have
looked through your posts several times, and I have read again those fro
others, in which you saw the attacks and denials. I am as puzzled by your
comments now as I was when I first saw them. Not one person on this net
ever attacked you, or said that social interactions, influences,
circumstances, and the like are unimportant.

Thank you, Tom. I'm glad someone noticed.

Ken, you have convinced me that I am correct in my decision to never again
submit a manuscript on interactive tracking and control to a real journal in
the social sciences. (I have one manuscript nearing completion on
"controlling behavior" -- taken both in the radical behavioral sense and in
the PCT sense -- that I will submit to a radical behavioral journal, but that
doesn't count as a social science journal.)

I would really like to get a copy of that paper; could you send me a
copy by snail mail or make it available to the net via the server?

Also, could you give a brief description of the two person cooperation
experiment that you presented at the Sociology Conference last year.
The people were connected to the lines in a most interesting way, as
I recall. In order to achieve their goals, each person had to, in a
sense, get the other person to "act properly". This was done by
disturbing one of the variables that the otehr person was controlling --
and vice versa. Perhaps you could just give the equations for the
connections between people and handles. Given this nice, quantitative
description of the nature of the connection between people in your
experiment, it might be easier for Ken to explain why this "card
stacking" tracking task tells us little about social interaction.

CHUCK TUCKER (930512) --

How do you know that someone understands PCT?

When they make demonstrably correct deductions and inferences based
on the PCT model.

I am wondering if one relieves himself of the
responsibility of influencing another more by using the
word "understanding" as contrasted with "convincing" another.

A person who is convinced can say only "PCT correct". A person
who understands says "PCT implies X and I can test that using procedure
Y". Such a person knows how to determine whether ANY deduction, X, is
correct (by running the model) and how to test the inference that
X will be seen in real systems (by using procedure Y). I can tell,
for example, that a person understands at least one aspect of PCT if
they know how to determine whether or not the model uses perceptual
information about the disturbance to determine the outputs that
compensate for the disturbance (once everyone agrees on the meaning
of these words in terms of variables in the model). Once you understand
PCT it's really not that hard to tell whether or not someone else does.

It was with my knowledge of PCT and my experience with you over
the years that I predicted (by private posts) that you and Rick
would commet on "boss reality" exactly as you and he did.

It was pretty exciting to me, too. Ah, to be in epistemological harmony
with William T. Powers.

But
someone with an S-R view of behavior would have predicted the
same reply by you and Rick.

Good for them; we're just two Rs in an S.

I was not playing games or trying to
trick you (or Rick)

Thank goodness! I would hate to have been tricked into revealing my
epistemological biases. I'm so shy.

Unfortunately, it spells trouble for PCT
in that it does not differentiate it EPISTEMOLOGICALLY from all
the conventional "theories" (including S-R or S-O-R) in the so-_
called "social and behavioral science." IT IS A REALIST (or NEO-
REALIST) THEORY IN IS EPISTEMOLOGICAL ASSUMPTIONS.

Actually, believing that there IS a boss reality is (as Bill noted)
quite a bit different than believing that you have direct access to
it. PCT is all about the fact that behavior is the control of a
perceptual construction of various aspect of boss reality. It is
perception that is important -- to both the actor and the person
studying the actor (the observer). PCT is quite clear in emphsizing the
fact that neither the actor nor the observer has priviledged access
to boss reality. That's why modelling is so important in PCT. And it
is modelling that sets PCT apart from most conventional "theories" of
behavior.

I would claim that is one of the reasons that reviewers do not see
appreciable differences between PCT and S-R; there is none epistem-
ologically.

I've never heard anything that even comes close to suggesting that the
reviewers have a problem with the PCT epistemology. The reviewers just
don't understand model-based science.

Best

RIck

From Tom Bourbon (930512.0123)

[From Rick Marken (930512.1000)]

Tom Bourbon (930511.1712)--

Ken, you have convinced me that I am correct in my decision to never again
submit a manuscript on interactive tracking and control to a real journal in
the social sciences. (I have one manuscript nearing completion on
"controlling behavior" -- taken both in the radical behavioral sense and in
the PCT sense -- that I will submit to a radical behavioral journal, but that
doesn't count as a social science journal.)

I would really like to get a copy of that paper; could you send me a
copy by snail mail or make it available to the net via the server?

As you have probably guessed by now, I am terribly slow about getting things
out. I work and re-work and start over and put away for a while -- not to
mention frequent lapses due to pure laziness. But I am forging ahead with
this one, having just revised the figures I that show the interference
task I used in Wayne's book. That is the task a person can turn into
"control of another," but "another" can then turn the fact of being
controlled into countercontrol of the one who controls ... yuk! Can you
believe radical behaviorists actually want to pursue those kinds of
interactions?

I promise you will have one of the very first near-complete drafts -- you
just volunteered to critique it for me.

Also, could you give a brief description of the two person cooperation
experiment that you presented at the Sociology Conference last year.
The people were connected to the lines in a most interesting way, as
I recall. In order to achieve their goals, each person had to, in a
sense, get the other person to "act properly". This was done by
disturbing one of the variables that the otehr person was controlling --
and vice versa. Perhaps you could just give the equations for the
connections between people and handles. Given this nice, quantitative
description of the nature of the connection between people in your
experiment, it might be easier for Ken to explain why this "card
stacking" tracking task tells us little about social interaction.

Only too happy to oblige, dude.

This is one I have been sitting on for over seven years. Maybe I should say
I have been "incubating" it for that long -- that is still sitting on it,
but it sounds much more intentional and productive. Actually, I haven't
been sitting on it. Three students did master's theses using this program
and a manuscript has been shot down a couple of times by those who know.

First, the environmental setup (no PCT here, just description), flipped on
its side because I don't know how to draw neat arrows that point up and down
and besides, trying to keep vertical lines aligned in a text editor gives me
the willies.

left person ----->left handle ---------->left cursor's position
                               >
                               >
                              \/
random disturbance--------->(ADD)-------->middle cursor's position
                              /\
                               >
                               >
right person----->right handle---------->right cursor's position

(See what I mean about those #*$&$ arrows?)

In this little world, the facts of life are that the position of the left
cursor (cl) is determined by the position of the left handle (hl), so that

             cl := hl.

The position of the right cursor (cr) is determined by the position of the
right handle (hr), so that

             cr := hr.

The position of the middle cursor (cm) is determined by the sum of the
positions of the two handles and the momentary value of a smoothed random
disturbance (d), so that

             cm := hl + hr + d.

And there was the evening and the morning of the second day.

****** Some of the possibilities in that world ******

If neither person moves a handle, the middle cursor (cm) moves up and down
on the screen, its position determined by d.

If either person moves one of the handles, what will happen? (This is a
test, not necessarily for Rick, but for anyone else who has read this far.
Before you read on, think of some of the options -- more than one.)

OK. Moving one of the handles affects two cursors, the one on the same
side as the handle, and the one in the middle. One person, using only one
handle, cannot keep those two cursors aligned, or in any other stable
relationship, on the screen. Moving the handle causes both cursors to move
-- unless. (Test #2: under which circumstance can the person move the handle
with the result that one of the cursors stands still? You might have
thought of this one during the first test.)

Got it? When the person decides to make the MIDDLE cursor remain at a
chosen place on the screen, then moves the handle to cancel the effect of
the disturbance. In that case, the handle, and the cursor it alone affects,
move to positions opposite and equal to those the disturbance would have
caused for the middle cursor, which remains at one place (+ - a small
variance) on the screen.

The other person can use the other handle to produce the same result.

How can two people, each using only one handle, keep the three cursors in a
chosen relationship (aligned, for example) on the screen? Acting alone, one
person can stabilize the middle cursor, so, for example, if the left person
acting alone decides to keep the middle cursor aligned with the right one
(which in that case is not moving), it can be done. Similarly for the
person on the right acting alone. Either person can achieve part of the
"superordinate goal" (a tip of the hat to Muzafer Sherif, and others), but
not all of it -- one person using one handle cannot create the intended
perception.

Test #3. How can two people create the perception of three little cursors
all in a row? ... Got it?

When people unfamiliar with this task first try to work it out, they often
assume they should try to keep "their own cursor" even with the other two,
but it can't be done. (They really do "personalize" those silly little marks
-- even I still do it.) If you leave people to their own resources on this,
all sorts of fasciating interactions break out right before your eyes --
but that isn't supposed to be happening -- after all, this is only
stick-wiggling.

The solution (a solution): each person can decide to keep the middle cursor
aligned with the cursor determined by the other person. But to do that,
each person must move a handle, which moves the cursor the other person is
using as a target for the position of the middle cursor, and the
disturbance independently affects the position of the middle cursor.
In linear language, this would never work -- or in linear models of
behavior. But with real people, it happens quickly and easily, once the
plan for perceptions is adopted. Each person controls a little (apparently
counter-intuitive) piece of the bigger pattern, which then emerges quite
nicely: the three cursors remain aligned, moving up and down on the screen
to positions opposite and egual to those where the middle cursor would have
gone if neither person acted. (Neither person intends to create that
effect, just as neither intends her of his actions. But both intend to see
the alignment of the three cursors, which they achieve by independently
controlling the relationship between two cursors. Have I said all of this
enough ways? It is much easier to just show someone, in real time.)

The two can select ANY ONE OF MANY other configurations, stable or
dynamic, for the three cursors, and create it quite easily.

We have used this task in several different setings: (1) partiicipants see
a demonstration, then practice until THEY think they are "ready;" (2) the
experimenter gives small bits of instruction or small hints, one at a time,
and the participants do their best, until, in some cases, they have been
told the entire "secret;" (3) participants are shown the computer, the
handles, the moving middle cursor, and left on their own to figure it all
out -- don't try this one unless you have a lot of time, or patient
participants; (4) the experimenter instructs one person on how to achieve
the overall pattern, then that person instructs the second participant, but
does not talk. I am sure you can think of many other variations on the
theme -- the task itself is merely a vehicle, for studying intraction, AND
for testing the PCT model of interacting control systems, which is next.
(Finally, the stuff Rick requested.)

The program steps that represent the environment are exactly as described
above:

     cl := hl; cr := hr; cm := hl + hr + d.

The PCT model for one person (on the left) must specify which variables, in
exactly which manner, cause the left handle to move, which affects the left
and middle cursors. The person's intended perception is represented by a
reference signal (rl) that in this case calls for: cr - cm = 0. The
person's perception of the relationship is represented by a perceptual
signal (pl) which quantity is the present value of: cr - cm. Remember, the
left person controls the middle cursor relative to the RIGHT cursor. The
position of the left handle (hl) is the sum of its previous position
(values of all variables are calculated to represent every 1/30th second of
a real run, so this would be the position from 1/30 sec ago) PLUS the
product of the estimated integration factor (kl) (which reflects
how far the real person moved the real handle when there was a given amount
of error) TIMES the error signal, which is the difference between
(cr - cm)intended [which is the reference signal, rl] and
(cr - cm)momentary [which is the perceptual signal, pl]. All of this is
expressed in the program as:

         hl(i+ 1) := hl(i) + kl * (rl - pl).

That is the PCT model of the person on the left.
By a similar line of reasoning, the model for the person on the right is:

        hr(i + 1) := hr(i) + kr * (rr -pr);

where reference signal (rr) calls for: cl - cm = 0; and where the
perceptual signal (pr) is the momentary value of: cl - cm.

The two extraordinarilly simple models, each of which controls its own
perceptual signal, produce results on the screen that closely resemble those
of the two people: correlations (I hear the groans!) between modeled and
actual positions of handles (two of them) and of cursors (three of them) are
all in the range of --- all together now: ____! And that is true whether
the models recreate a previous run, or predict a later one. On this
evidence, I rest my case that two people can be MODELED AS THOUGH they were
two independent control systems, each controlling its own perceptual signal
and neither alone capable of producing the overall effect they create
together. I do not know how to model the speaking, gesturing, jumping up
and down (literally) and other communicative acts that can occur while two
people work to get to the point where I model them this way, but those
acts are deliberate, purposive and indicative of control. I do know that
the communications that work best are those about what the perceived
results will be when "all is right" and the least effective are those in
which one person tries to inform the other about specific actions and
movements.

One person can perform the task using both hands; two PCT models can perform
the task together; a person and a PCT model can perform it together. One
person can use a mouse and one hand -- all of the permutations I mentioned in
the post on the interference task.

This is the work that prominent experimental social psychologists do not
want to see. I wish them well.
Until later,
  Tom Bourbon