[From Bruce Nevin (980520.1610 EDT)]
Rick Marken (980519.2100)--
It only matters whether the coffee maker is a control system
controlling for keeping the coffee inside itself.
Tim Carey (980520.1035) --
[...] you and Bill maintained that coercion was when someone
controlled the other person _regardless_ of the references of the
other person.
[...] The way I said it above it seems like I am
saying that coercion occurs only when the coerced control system
(coffee maker) wants a perceptual variable (location of coffee)
in a _different_ state (inside itself) than that desired by the
coercer control system (outside). In order to prevent this
misconception, I should have said:
It only matters whether the coffee maker is a control system
controlling for the same perceptual variable (location of the
coffee) as you (the coercer).
It sounds like coercion is simply a limiting case of conflict where control
by one control system overwhelms control by the other. However, I have some
methodological and conceptual problems with this account.
The phrase "same perceptual variable" glosses over the issue of what it is
for perceptual variables in two different control systems to be the same.
The place where they are the same is some physical variable X in the
environment. We have only our perceptions of X, but the actual point at
which the two control systems conflict is not in their perceptions of X, it
is in X itself, whatever it is. For purposes of modelling, it is an
attribute of something in the environment as measured and quantified by an
observer.
As observer, it is precisely the conflict between A and B over X that
enables us to conclude that some perception A(x) that A is controlling is a
perception of the same environmental X as some perception B(x) that B is
controlling. (Cooperation over X, e.g. verbal agreement about it or
interacting consistently with one another over it, is another basis for
concluding that we are perceiving the same X.) We assume that X is actually
what is given by O(x), the observer's perception of X. But to say that A
and B are controlling the same perceptual variable is not possible even in
a model.
Look at a diagram:
>
> r
[CS A] V
+----->COMP ----> e
> >
p = A(X) | |
> >
- - - - - - Fi - - - - - -Fo - - - - - -
^ |
> >o(A)
> ___ |
+------| |<-----+
d--------------->| X |-------------+ [ENVIRONMENT]
+--------------|___| |
> ^ |
p = B(X) | | o(B) |
- - - - Fi - - - | - - Fi - F0 - - -
> ` | ' | ^
> ` | ' p = O(X)| |
V ` | ' V |
COMP----e--->Fo--+ ' COMP----e
^ ' ' ^
> r ' ' | r
> ' ' |
[CS B] ' ' [OBSERVER]
' '
' '
There is no way, even in a model, to say that the three perceptual input
signals A(X), B(X) and O(X) are the same. You are talking about an
environmental variable X as measured from the observer's perceptual input
O(x). OK, an environment variable that the two control systems want to
perceive as having different values. But there's more than just careless
use of terms that bothers me here.
Suppose A wants the door shut because it's supposed to be shut and B wants
to open it enough to let in a breeze. Already, X is different for A and B.
For A, X is the doorway being filled up and closed off with the door; for
B, X is the feeling and smell of the spring breeze. By asking and Testing
we can determine that they are in fact not controlling the same perceptual
variable at all. This of course is the basis for finding win-win solutions
to conflicts, instead of e.g. assuming that the bone of contention is the
physical locus of conflict (position of door, 20 degrees vs. 0 degrees).
The observer can quantify their conflict in terms of the position of the
door because some X in the environment that the observer perceives as the
position of the door is the place where the effects of B's efforts and the
effects of A's efforts come in contact in the environment.
The earlier claim is that the coercer is controlling an aspect of the
behavior of the coerced system. Suppose that A is coercing B. The claim has
been understood to mean that A has a reference perception for B's
behavioral outputs, o(B).
>
> r
[CS A] V
+----->COMP ----> e
> >
p = A(X) | |
> >
- - - - - - Fi - - - - - -Fo - - - - - -
^ |
> >
> ___ |
> > > >
d---------)----->| X |------)------+ [ENVIRONMENT]
+-------)------|___| | |
> > ^ | |
p = B(X) | +-------)-o(B)<---+ |
- - - - Fi - - - | - - Fi - F0 - - -
> ` | ' | ^
> ` | ' p = O(X)| |
V ` | ' V |
COMP----e--->Fo--+ ' COMP----e
^ ' ' ^
> r ' ' | r
> ' ' |
[CS B] ' ' [OBSERVER]
In this diagram, A is controlling a perception of some observable aspect of
o(B) such as B's arm position, while B is controlling a perception of X.
(The observer is presumably controlling perceptions of o(B) and o(A)
without any error output. I have not shown that. There's a question about
this at the end.)
Of course, what they are controlling might change. In particular, B might
shift to controlling some aspect of o(A), I suppose as part of a sequence
perception of which controlling a perception of X might be a later step.
A might also have a reference perception for X, so that A and B are in
conflict over X. In particular, A might require o(B) as a "tool" for
controlling a perception of X.
Rick, your claim was that the effect of o(B) in the environment is an
aspect of the behavior of B. In this way, you could say that the coercer,
A, is controlling an aspect of the behavior of the coerced system, B, even
when A is only controlling a perception of X but is overpowering B's
control of a perception of X. I think this meaning for the words "an aspect
of behavior" -- meaning any variable in B's control loop -- has been
confusing.
I suppose any irresistable disturbance to B's control of B(X) that is due
to o(A) constitutes coercion of B by A. The disturbance might be
1. An environmental effect of o(A) on qi and thence on B(X) that the
observer sees as A's effect on X.
2. An environmental effect of o(A) on qi and thence on B(X) that the
observer sees as A's effect on B's ability to sense the state of X (putting
a blindfold on B, injecting B with novocaine, turning out the light).
3. An environmental effect of o(A) on qo as it results from o(B) that the
observer sees as A's effect on B's ability to affect the state of X
(holding B's arm).
4. A's depriving B of some environmental contingency such as a tool that B
would use to affect the state of X.
This broad definition does seem to follow from understanding behavior to be
the control of perception, and coercion as affecting not just behavioral
outputs such as o(B) or those aspects of o(B) that an observer perceives.
Under this definition, your demo does model coercion where o(A) affects the
same environmental value X that o(B) is intended to affect but cannot. You
know it is the same value because it is a product of the program that you
wrote. Mostly that cannot be said of environmental variables.
Saying that the coercer and the coerced control the same perceptual
variable is just sloppy talk. They can't.
Imagine now coercion situation in a typical school. John, a teacher, is
controlling perceptions of his students learning foo. Freddy is controlling
a perception of talking to Alan, another student. This does not satisfy
John's controlled perception of Freddy learning foo. So far we have
different perceptions being controlled, learning foo and talking to Alan.
We could press down through the hierarchy looking at the means for talking
to Alan vs. the means for learning foo until we found some particular
variables for which John and Freddy are controlling different values. But I
think it's more accurate to say that they are controlling different
perceptions entirely. The teacher wants Freddy to control a perception of
learning XYZ at the moment (and of learning in school in general). The
conflict is over what perceptions Freddy should control. The coercer and
the coerced do not have different values for the same variable (as
identified by an observer) unless you search down through the hierarchy to
the particular point where John's means of coercing Freddy actually or
potentially come in contact, in the environment, with Freddy's means of
controlling a perception of talking to Alan.
As I understand it, in an RTP school John the teacher is controlling a
perception of none of his students being prevented from learning XYZ.
Freddy can be controlling some perception other than learning XYZ so long
as his doing so does not interfere with any other student learning XYZ. If
any student pays attention to what Freddy is doing, it's a distraction. If
any student is unable to see or hear John, it's a disruption. The conflict,
I expect, is usually about unintended side effects of Freddy's controlling
something. It could even be a side effect of Freddy's way of controlling
learning XYZ (talking to himself, say). Of course Freddy could be
controlling a perception of disrupting John's teaching or of somebody's
learning, and as noted this could be as means of controlling a perception
of getting out of class to the RTP room. I just suspect that that is less
common than unintended side effects. This is consistent with my experience
that growing up involves getting better at managing side effects of
control. But in the RTP situation it is no easier to make out the coercer
and the coerced controlling the same variable in conflict, especially if
the conflict is over unintended side effects.
To model conflict and coercion we have to get beyond the case of two
control systems controlling the same perceptual variable with different
values. It is too simplistic. The modeller wants a single variable at the
crux of conflict. That's like Nasrudin and his house key.
Nasrudin's friend found him crawling around the sidewalk under the
streetlamp. "What are you doing?"
"I'm looking for my house key."
"I'll help you."
After half an hour, the friend grew impatient.
"Are you sure you lost it here? We've scoured every inch!"
"No. I lost it over there in that dark alley."
"Then what are we looking here for?!"
"There's more light here!"
We might say that the Observer's lack of output o(O) is due to zero gain,
but we have suggested zero gain as an index of zero attention. I know
you've said something about this, Bill, but my head is still thick about
it. As a guess a copy of perceptual input might be used as the value of r
so there's never any error, a kind of corollary of the imagination loop
perhaps used for learning or memorizing, but there's probably a simpler
mechanism I'm just not getting.
Bruce Nevin