[From Bill Powers (920901.1000)]
Dag Forssell (920831) --
What happened here? Otherwise looks good.
If the associate wants the shareholder........ Associate
satisfaction is all there is. Associates is all you have to work with.
Associates must want and perceive many things and take effective action
for the company to function well.
···
-----------------------------------------------------------------------
Martin Taylor (920831) --
... when we are trying to be precise, and communicate exactly, then we >have to
go back to separating the controlled percept from the CEV, >which is what I was
trying to do. We do, easily but sloppily, use >third-stage language most of
the time.
But doesn't this say that there "is" a CEV with its own definition Out There in
the environment, and that the percept simply represents it?
I'm trying to make the percept the central concept, because that's what the
control system knows and what it controls. The CEV is found (by somebody who
knows more than we do) by working backward through the perceptual function and
finding what combinations of external variables would be sufficient to produce
that percept via that input function. In general there are many combinations of
external variables that will lead to the same percept -- that's the other side
of the concept "invariant."
We have to distinguish between a CEV as seen by the controlling system and as
seen by a different person, a different perceptual hierarchy. This will not in
general be the same CEV. The only completely correct
-1
definition of a CEV is f (p), where f is the input function in the controlling
ECS. Because the inverse of a many-to-one function is always ambiguous or
multiple-valued, all we can say about the CEV as deduced from the inverse is
that it consists of ANY ONE of the combinations that will yield that value of
that percept. We have to consider many different CEVs in order to pin down the
environment to a less ambiguous definition of each CEV. So a single CEV really
doesn't exist as a single independent unambiguous entity when considered alone.
Through joint efforts like physics, we can come to agree on definitions for
certain CEVs like mass, energy, shape, color, and so on. But for each of us,
these definitions remain private; we can't go further than being satisfied that
other people seem to use the agreed-on terms in situations that we find
consistent with our own perceptual meanings of the terms. What a given person
controls when controlling for "energy" is precisely the perception to which
this word points. But that perception is known only to that person.
As you put it here, ALL CEVs are and always will remain misdefined, >inasmuch
as there probably exists no construct whatever in the outer >world that behaves
precisely as the perceptual input function that >defines the CEV does.
Yes, that's another way to say it. But you can also say that for each person,
the CEVs are precisely defined, but coincide only in part with the the CEVs
that other people also precisely define. The ideal of defining a CEV so that
all people will agree (operationally) is not reachable. And of course the idea
that there are objective CEVs independent of our ways of perceiving them is a
question we would all like to have answered. But there seems to be, at present,
no hope of answering it.
Your concept of X-control seems to be just a way of desribing the control
process that one person attributes to another. The Test reveals X-control. We
hope it bears some resemblance to P-control, and keep looking for ways to make
that hope more realistic.
RE: transients.
You can't have it both ways. Either there is an error signal that >results
from the disturbance and generates an output signal that >resists the
disturbance, or there isn't.
We're having a quantitative misunderstanding here, resulting from using words in
a qualitative way. There are degrees of error and degrees of effect of a
disturbance on a variable. If the variable is under control, a disturbance will
affect it a lot less than if the same variable were not under control. I'm
drawing the line between "effect" and "no effect" between these two cases.
In a control system with a loop gain factor of -30, a disturbance that would
have 100 units of effect on the variable when uncontrolled will have only 3
units of effect when control is present. So yes, the disturbance does disturb
the variable, but in comparison with the effect that would occur with no
control the amount of effect is essentially nothing. It is, in fact, so small
that it can be neglected. If that were not true, the control system would have
continued to reorganize until it had a gain of 300, or 3000, or whatever was
required to make the deviations unimportant to it.
And if the transient response delay of the control system is 1 second, then
disturbances that last less than that amount of time will be unimportant to the
control system. If that were not true, the control system would have evolved to
respond in 1/2 second, or 1/4 or 1/8 second, whatever is required (within the
limits of the possible).
When I spoke of a transient effect on the controlled variable, I meant an
essentially _uncontrolled_ effect -- in other words, an effect that is just as
large as it would have been with no control. In a control system whose output
rise time is 0.2 seconds, a disturbance that lasts only 0.05 seconds will alter
the variable just as if no control existed. But the control system will not
respond to that change, so that brief a disturbance can't be used to control
the action of the system. You must maintain the disturbance long enough for the
output to come to equilbrium with it if you want to control the output.
You understand me correctly in saying "It means that the organism will
survive, behaving this well, and that no excess resources will be
applied to ensure better behaviour than that."
We also agree, I think, about manipulation:
No difference when we "control" another person. We keep trying
different "influences" until the person does what we want or until we >give up.
Right. Manipulation, if goal-directed, always comes down to control. And when
you try to control another control system IN ANY RESPECT THAT IS ALREADY UNDER
CONTROL BY THAT SYSTEM, you generate conflict, a contest of brute strength.
Such contests can be fun when done for the purpose of developing skills. And
when you know the game will be over pretty soon. As a way of life, they stink.
No, control of another has nothing whatever to do with setting >reference
levels for another. Control of another may be achieved >easily if it so
happens that the other's reference levels come to lead >to actions that are
perceived as appropriate to the reference levels in >the "controller." But the
other's reference levels are a red herring >in the discussion of whether one
person can control another.
Right on. Either you control other people in the trivial sense that you elicit
error-correcting actions from them that happen to suit your own purposes,
without serious inconvience to the other, or you wrestle with them over who
gets to control something that can't be in two states at once. So far our
civilizations have been founded on the latter mode of interaction, which
accounts for their violent history.
I do want to cavil about control existing when one person happens to control
something the way somebody else wants it controlled. I said this to Greg
Williams but it's worth repeating. Control requires a closed loop. If one
person is controlling a variable, it's highly unlikely that another can control
it at the same time without generating conflict. So the spectator may be
satisfied with how the other person is doing the controlling (as the spectator
sees the CEV), but the spectator can't DO anything about it if an error
appears. The spectator is not controlling, because controlling requires action
to close the loop. The passenger can jam down on an imaginary brake, but had
better not jam down on the real one if conflict with the driver is to be
avoided.
------------------------------------------------------------------------
Mark Olson (920831) --
Tell your friend to hang on; I'm working on an extensive writeup for Little Man
Version 2. I'll announce it when it's done. I'm going to sell this program for
$100, but have decided to give an 80% discount to CSG members (Dag note). And
of course it's basically shareware because people will rip it off anyway. I
don't really mind if people who can't afford it or can't finagle department
funds just copy it. But I am trying to pay for my new computer, and want to
build some experimental apparatus, and get Frans Plooij to the next meeting,
and all that sort of stuff, so the money is really needed.
----------------------------------------------------------------------
Martin Taylor (920831.1720) --
"All behaviour is the control of perception." But is it?
Rick said it: all control behavior is control of perception. The problem here is
again language. "Behavior" can mean anything we see happening anywhere, in
organisms or otherwise. The behavior of a rolling marble. The behavior of a
thunderstorm. It can also mean irrelevant side-effects, especially if you don't
know anything about control theory.
The problem is that historically there's been no understanding that the behavior
of an organism while controlling something is different from any other sort of
activity we might notice. In one sense, behavior has been used to refer to
events almost as if they were intentional when in fact they aren't. To speak of
the behavior of a marble rolling around in a bowl carries a subtle idea that
the marble has something to say about how it's going to move -- we attribute
the behavior to the marble almost as if the marble is an agent. In fact the
path of the marble is due entirely to external forces acting on its spherical
shape and mass which are the same no matter what the marble is "doing."
Exploratory behaviour is "what if" behaviour. One is not controlling >for any
particular reference (at a higher level one is controlling for >a percept of
knowing more than one does, but that's outside this >point). One is just
acting, more or less randomly at the level at >which exploration is being done.
Of course it isn't random at lower >levels, and probably not at higher ones.
Exploratory _behavior_ -- meaning observable actions -- is just like any other
kind of behavior. If you see someone sharpening a pencil, you can't tell
whether he's sharpening it because he doesn't know what will happen and wants
to find out, or because he does know what will happen.
What's different about exploratory behavior is the goal, not the behavior. At
some level, as you say, the action must differ from ordinary action, in that
it's more or less random (although when you do a search for a lost child, you
organize to do systematic sweeps of the territory, so all that's exploratory
about this is not knowing whether the error will be corrected or not).
The goal-directedness of exploratory behavior becomes obvious when you ask what
will make it stop.
--------------------------------------------------------------------
In the Layered Protocol formulation, before I heard of PCT, I argued >that the
purpose of casual conversation was for the partners to develop >models of each
other that would permit them to communicate effectively >when the need arose.
In the same vein, I suggest that the purpose of >casual exploration is to
develop organizations that permit effective >control when the occasion arises.
Exploratory behaviour, then, is NOT >the control of perception, but the
discovery of how to control >perception.
Tsk, tsk, Martin. Are you saying that the goal is to learn how to control, but
without any perception of whether you've learned to control? You forgot to go
up one more level.
---------------------------------------------------------------------
Jeff Hunter (920831) --
Thanks for the further info on the Toast approach.
Planning certainly happens. If people do it, it belongs in a model of people.
The only question I have is how elaborate it really is, and at what level it is
really applied. From what I've seen of planning programs, they include a lot of
things from lower levels of organization that could be done a lot more simply
just by control systems. By confining the planning aspect of the model just to
the levels where planning alone is the problem, you can make the planning tasks
much easier and also do a better job of explaining behaviors that really don't
have any planning aspect to them. In a hierarchical model you can take care of
lots of things, like controlling spatial relationships, without getting into
classifications, sequencing, and logic. That means that the planning system can
quit worrying about the details, and simply solve conflict problems, perhaps
without even having to know what the conflicting variables mean. One variable
can't be in two states at once; that's all the planning
system has to know.
----------------------------------------------------------------------
Ray Jackson (920831) --
Welcome aboard, Ray! Be patient; the threads will intertwine and sooner or later
the one you want will come along. Feel free to partipate in anything, and don't
hesitate to raise questions of interest to you.
-----------------------------------------------------------------------
Avery Andrews (920901) --
I'll wait until you get the post I just sent. Try again. I'm a little swamped.
--------------------------------------------------------------------
Best to all,
Bill P.