Attn. P. Sibun; repentance (might the end be near?!?!)

From Greg Williams (920925)



Penni Sibun: please post your USPS address so Ed Ford can send you a sample
copy of CLOSED LOOP.


Bill Powers (920924.0900)

[in reply to Greg & Pat]

You don't understand yet. From Pat's point of view they are also not
acting differently.... "Action" means what they are doing with their muscles.
She hasn't changed that. She's only changed an effect of the actions in a
dimension that leaves the control systems undisturbed, doing exactly what they
were doing before.

You're right. I was wrong. This example of interaction is NOT analogous to
rubber-banding. Pat is being careful NOT to disturb the kids' controlled
perceptions (of relevance here) in order to control certain of her own
perceptions which depend on certain of the kids' actions. So, there are at
least TWO kinds of "short-term" purposive influence:

  1. By disturbing the second party's control of some perception in such a way
that the second party performs actions (to compensate for the disturbance)
which enable control by the first party. ("Rubber-banding")

  2. By altering the environment of the second party in such a way that the
second party's ongoing control of perceptions is not disturbed, and that
the second party's ongoing control (given the altered environment) enables
control by the first party. ("Food-spiking"?)

For either of these types of interactions to be successful from the point of
view of the first party (that is, his/her control is good), it is important
that the first party have a good model of what the second party is controlling
for/not controlling for -- perhaps more to the point, a good PREDICTIVE model
of what the second party WILL control for/not control for WHEN the first party
disturbs (type 1 interaction) or alters the environment so as not to disturb
(type 2 interaction). Applying The Test (in some form, perhaps repeatedly) is
key to developing and refining such a model. Empirically, I see both types of
interactions occurring frequently, with widespread (but not universal) success
from the standpoint of the influencers, and (afterwards) characterized as
ranging from "good" to "bad" for the influencees, as judged by the
influencees. Your earlier comment that the second kind of "short-term"
purposive influence is barely an interaction at all appears to discount the
model-building aspect (applying The Test), which is usually HIGHLY

All right, so (despite my density!) we've made some progress. It appears that
we've discovered a kind of interaction wherein party A can (in theory, at
least) control certain of his/her perceptions which depend on another
organism's controlling and which is not of the rubber-banding type or of the
(hypothetical) "guiding" learning/reorganization type. (You might recall
that, a few weeks back, you thought it best to lump deceptive "short-term"
purposive influence with rubber-banding, because both involve applying
disturbances, even though in somewhat different ways.)

Whether or not you are interested in such "short-term" purposive influencing,
I'd appreciate your remarks on the above conclusion. I'd hate to set out on
the wrong foot again!

[in reply to Greg]

You have shown that it's possible for one person to control the actions of
another person by disturbing the variable that the action is involved in

Well, I THOUGHT the Pat-feeding-"healthy"-food example showed that. But, as
you say, I was wrong. What I really showed was that it's possible for one
person to control his/her perceptions by being careful to NOT disturb
another's controlled variables, while altering the other's environment in
certain ways. YOU showed, long ago, by inventing the rubber-banding demo, that
"it's possible for one person to control the actions of another person by
disturbing the variable that the action is involved in controlling."

You have cited, as a way of overriding my objections to your argument, the
rubber-band experiment and the particular use of it that shows how one person
can control another person's finger position -- both of which demonstrations I
invented and used to illustrate exactly the same point.

Yes I did, wrongly. The Pat... example is NOT analogous to rubber-banding,
which is your brain-child. Actually, the Pat... example is of a completely
different sort than rubber-banding.

Unfortunately, you chose as an example a case in which this demonstration
doesn't apply.

Yes, I hear you. At least my mistake was not all for nought.

This suggests that you believed in the conclusion, but didn't understand how
it was reached. Of course I know that you DO understand how it was reached.

Honestly, I was confused by the "low-level mechanics" of the kids' actions
being different if Pat was there vs. if she wasn't.

I can only conclude that you thought it more important to prove that you were
right than to do so by a rigorous argument.

I thought it was rigorous at the time. I did not and do not think it more
important to prove that I am right than to argue (as best I am able)
rigorously. Perhaps there is no way to convince you of that. Nevertheless, I
will say explicitly that I am telling the truth.

The outcome was that you have slipped and shifted and misstated, flailing
around with the sole apparent purpose of convincing me that situations
analogous to the rubber-band experiment work exactly as I say they do. I
agree. You win. These relationships work exactly as I believe, and have
believed for a long time, that they do.

I greatly admire your many accomplishments, including inventing the rubber-
banding demo, of course. Far be it from me to not give credit where it is due.
And now you have also participated in the discovery that PCT shows there is
the theoretical possibility of DIFFERENT kind of controlling-dependent-on-
others'-controlling IN ADDITION to your rubber-banding kind. As I said above,
I think that's real progress in using PCT to understand the nature and limits
of social interactions, and I appreciate your patience, especially since say
you aren't very interested in (as I call it) "short-term" purposive

I hope to extend this work, further considering the constraints on social
interactions which follow from PCT. I know that you are interested in talking
about what I've been calling "long-term" purposive influencing. I'd also
welcome discussion with others who are interested in social interactions in
general, whether they involve learning/reorganization or not. Wherever we go
from here, thanks, Bill, for helping me get this far!

Guided reorganization aside ...

You'd better explain how you can guide a random process first.

There are random processes and there are random processes. Some kinds of
"stochastic" random processes combine random aspects with non-random aspects
so that you can predict where they are headed, but not the trajectory they
will take or how long they will take to get there.

As a metaphor, suppose that a ball with several motors distributed over its
surface (not unlike the hairs on microorganisms) is in water in a swimming
pool. Each motor propels the ball along a particular direction with respect to
a ball-centered coordinate system (the ball doesn't float), and each motor
turns on (and, simultaneously, all others turn off) randomly in time, for
random-length time intervals. At one end of the pool is a hole through which
the ball could pass, say, to ball-nirvana. The ball will either eventually
move through the hole and reach ball-nirvana, or never (in infinite time) move
through the hole. In this simple case, there is only one hole to move through,
no others.

If the ball's random motion is analogous to learning/reorganizing and going
through the hole (conveniently provided by the God of Balls -- an outside
influencer) is analogous to learning/reorganization ceasing (due to finding
the one solution to the problem set by the "guider"), there's your answer. IF
the solution is found, it is the "guided" one. Of course, the solution might
never be found -- possibly because the control system doesn't know the
Pythagorean Theorem. Solution: More "guiding," through intermediate stages of
learning/reorganization. Please don't presume I am implying that there
necessarily must be ONLY ONE solution, or even a FEW solutions to the problem,
set by the influencer. The influencer might be happy to see any of a HUGE
CLASS of solutions. And please don't presume that the influencer must want to
seeing the solution reached in a highly particular way; the influencer might
care not at all about all of the trial reorganization attempts, and value only
the eventual success -- perhaps just seeing the number "6," regardless of the
manipulations the influencee went through to write it.

In such a case, the obvious strategy for the "guider" who wants "tight"
control of his/her perceptions of where learning/reorganization is "taking"
the influencee's control system is to break up the whole process into little
"nearly obvious" steps. "Nearly obvious" means that the (class of) "new"
outputs of the influencee which the influencer wants to see after a bout of
learning/reorganization has finished are not very "far" is modeled by the
influencer as requiring little alteration in the influencee's control system.
For example, the influencer doesn't expect to successfully "guide" the
influencee, in one bout of learning/reorganization, from competency in
counting to competency in long division.

Influencing so as to "guide" learning/reorganization is, no doubt, an art
which is not guaranteed to succeed. But it doesn't differ in principle from
control of inanimate entities, which also requires applying a model (granted,
very crude models are sometimes sufficient here -- like "this rock will do the
same thing as it did last time under these sorts of conditions"); the
difference is that good control of one's perceptions which depend on someone
else's outputs after learning/reorganization requires, in general, fairly
sophisticated modeling of the other's controlling before and after
learning/reorganization, in order to alter the other's environment so each
step in the process is "nearly obvious" to the influencee. Another metaphor:
the "guider" casts a sufficiently large net that the random floundering of the
fish won't carry the fish outside the net, and then the "guider" makes sure
(for each "nearly obvious" step in the learning/reorganization process) that
the net has one outlet into a similarly arranged next net. Modeling is key to
deciding how large the net must be.

It seems to me that "guiding" learning/reorganization would be made easier if
the learning/reorganization process does NOT proceed by making LARGE "jumps"
in particular reference signals, but by making SMALL deviations. If, during
learning/reorganization, the values of reference signals resulting after one
"trial" alteration of reference signals had absolutely NO correlation with the
values before, then it might be EXTREMELY difficult for "guiding" (in
particular, through "nearly obvious" stages) to work. But if such a lack of
correlation INVARIABLY characterized learning/reorganization, it seems to me
that successful learning/reorganization would almost always never happen. So I
suspect that much -- though not necessarily ALL -- of the time,
learning/reorganization is NOT characterized by such a lack of correlation.
Perhaps unless success has been a VERY long time in coming via a series of
successive small deviations in (sets of) reference signals, large jumps in
reference signals aren't made.

All this is highly hypothetical, of course. But, so far, I see no internal
inconsistencies. But I hope you will enlighten me about where, if any place, I
am going wrong.

... I claim that purposive influence is intended (by the
influencer) to have ONE effect, namely enabling the influencer to
control some of his/her perceptions which depend on some of the
influencee's actions.

... I say balderdash. If that's all that purposive influence amounted to, we
wouldn't even need a name for it.

I don't understand why you say this. To me, it seems like a pretty big "all";
one that sociologists, among many others besides the would-be purposive
influencers and influencees themselves, are greatly interested in.

In fact, purposive influence even in the rubber band experiment is
specifically aimed at controlling _a perception of the other person's action_.
The point is not just to control some other perception that's dependent on
those actions; the point is to make those actions, as perceived, be exactly
what you want them to be.

I didn't say "some other perception" than the perception of the other person's
action. I said "some... perceptions which depend on some of the influencee's
actions." Isn't the perception of the other person's action a member of the
class of perceptions which depend on the other person's actions? My intent was
to not limit the possible controlled perceptions of the purposive influencer
to only the perception of the influencee's actions. My definition INCLUDES
what you say "amounts to" MORE than my definition! I think what you really
mean is that you want to RESTRICT the definition to controlling ONLY
perceptions of influencees' actions. Right? Assuming so, then consider the

I tried to point out the difference in your passing-the-salt example. If all
I want is to control a perception that depends on some of your actions -- in
other words, if all I want is to perceive the salt shaker in my hand -- then
I don't care who passes me the salt or how. I am not controlling the action
by which you pass me the salt. It can vary all over the place and I will do
nothing to restore it to any particular form of passing the salt. What I'm
controlling for is the salt, and perception of your action in passing it to
me is not part of my controlled variable.

This is a BIG difference. If I say "Please pass me the salt," you not
only are free to comply or not, but you can achieve the result of
passing the salt in any way that's convenient to you, including asking
someone closer to me to do it instead. But if I want to have a
PURPOSEFUL influence on your ACTION, I will say "Please move your hand
to the salt shaker, grasp it, lift it six inches, move it in a
straight line to a point over my hand, lower it, and let go." That is
purposeful influence of your ACTION, if it works.

It appears that these two types of relationships between the influencer's
controlled perceptions and the influencee's actions correspond to the two
kinds of purposeful influence (1 and 2). In rubber-banding, as you say,
"purposive influence... is specifically aimed [by disturbing the influencee's
controlled perception in certain ways] at controlling _a perception of the
person's action_." But in food-spiking, purposive influence is aimed at
controlling a perception which depends on some of the influencee's actions,
but does not involve having a purposeful influence on the influencee's actions
[it involves altering the influencee's environment, taking into account (a
model of) the influencee's controlling so as to NOT alter the influencee's

Do you have any problems with this reformulation?

The influencer controls by disturbing what the influencee is
controlling for (or however you want to name the "rubber- banding
interaction" discussed above), without force or threat of force,
and without trying to intentionally change what the influencee is
controlling for. Is that "control of others"?

Yes. It's control of an action by another. No controlled variable in
the other person is disturbed significantly, meaning to an extent that
the other can't easily oppose without inconvenient effort.

Rubber-banding-type interactions are "control of others' actions," but food-
spiking-type interactions are not. Still, of course, both afford party A a
means of controlling certain of A's perceptions which depend on B's
controlling. But only in the former case are disturbances applied deliberately
by party A (those disturbances being effectively counteracted by party B's
good controlling, so that party A sees the actions of B which party A desires
to see). In the latter case, non-disturbing changes in B's environment are
made by A so that when B controls in the absence of disturbances, A sees the
perception A wants to see, which depends on B's actions, which are NOT
controlled by A. Whew! I think I've got it at last.

It's not "bad" or "good" when it works as intended. It has no effect
on any controlled variable in the other.

Of course, rubber-banding does have SLIGHT effects on some controlled
perceptions, unless control is PERFECT. "Food-spiking," if done well from the
standpoint of the influencer, has NO effect. Right?

When the action you're making the other perform has a side-effect of
disturbing some other variable that's under control, the other will resist
the side-effect. If that resistance is successful, there's no problem. If,
however, a very large effort is required in order to counteract the side-
effect (for example, performing the action may result in fatigue), then the
purposive influence will most likely fail. If the purposive influencer doesn't
change the purpose, he or she will simply try harder. This will lead to
conflict. If the purposive influencer considers conflict "bad", the only
solution is to change the purpose.

Of course, that goes for rubber-banding-type interactions only; not for food-
spiking-type interactions, in which you DON'T "make the other perform" a
certain action. Right?

Skinner thought that teaching and psychotherapy amounted to nothing more than
getting organisms to produce particular ACTIONS.

Yes. That is what he called "controlling" the organisms. At least I'm not THAT

My aim now and for the past month has been to correct what I perceive as
errors in the application of PCT principles. I state my corrections. What you
do with them is up to you.

I hope I am doing justice to your aim. One thing is for sure, whatever you
might make of it: I am treating your ideas seriously.

Thanks again for your patience, and especially for providing explicit answers
to many of my questions in my last post. I promise not to post the back log!!

I'm off to the Great University of Kentucky Book Sale, surpassed locally only
by the Great Dayton Planned Parenthood Book Sale. Greg-nirvana! See you


P.S. Is the arm program all fixed? How's the paper rewrite coming?