adaptive control;teaching

From Tom Bourbon [930907.1204]

[Hans Blom, 930907]

To Bill Powers, Rick Marken and Tom Bourbon

..

Tom, thank you for your point of view. I need to take more time for your
extended reply than I can spare at the moment. Just a short remark, how-
ever: adaptation in a control system can be realized in a multitude of
ways. Sometimes one method is better than another, sometimes it is the
other way around. Much depends on the stability of the "outside" world,
that the organism interacts with, and on its orderliness. It is very
difficult, for instance, to find the highest peak in an Alpine landscape
if you do not have eyes (long range sensors). It is even more difficult
to find the highest peak on earth.

Concerning the problem of locating the highest mountain peak *anywhere*, I
agree that the task could be difficult for an organism without eyes, or a
functional equivalent for detecting the relative elevations of places both
near and far. On the other hand, I imagine very few individuals who are
members of species that lack those sensory devices ever conceive of and try
to locate "highest mountain peaks." *We* recognize this as a possible
"problem" for an artificial device, but only because we have experience with
"maxima" and with having been "tricked" by a local maximum.

I agree that there might be more than one way to achieve adaptive
control. My intention, in my earlier reply to your reply to Rick, was
merely to describe a model for adaptive control that, so far as I can tell,
assumes all steps in the process can be modeled in terms of the fundamental
ideas in PCT. That model worked in the simple application I explored. I
would never pretend that one demonstration under those conditions implied
*anything* about the functionality of other schemes for achieving adaptive
control. On the other hand, the demonstration I described does serve as a
refutation of claims or suggestions (not from you) that PCT models are
static and unchanging and therefore cannot serve as models of living
systems. It was possible for us to refute those allegations prior to the
demonstration of adpative control by a PCT model, but the refutation can be
stronger as a consequence of the demonstration.

Also, the fact that a random element can be built into adaptation does not
rule out non-random adaptation, perhaps involving formal rules. The latter
was the first type of adaptation I used in my modelin of one person
controlling the actions of another, and it is still the approach I use when
I interact with another person in one of my two-person tasks. I already
know the ways a control devices affect variables on the computer screen so I can
use rule-based strategies when I "adapt" my actions to an unteraction with
a particular new person. The random (E. coli) model was a test of whether
an untutored or inexperienced control system, with no formal rules and no
bases for assuming anything about which "direction" to change, could achieve
control. That model could serve for either (a) a "simple" system or
organism that lacks the "hardware" for ever developing plans or strategies,
or (b) a system-organism with the requisite complexity, but with little or
no experience controlling under particular conditions. In the former case,
the random procedure might be the only one that can ever occur; in the
later, it might be a mechanism (perhaps the only one, perhaps one of
several) for reorganization (ie, "learning" or "development").

Just a short question. Do you think that, physiologically, it might be
possible that effector information might be available to higher levels of
the nervous system? Does the information that the brain has available
include what I call the output of the organism, i.e. its actions? If not,
physiology would rule against my "input-output" identification scheme.

Bill Powers (930907.0830) replied in the affirmative concerning some
sensed variables (muscle length, angles of bones around joints). To that
list you can add accelerations (including zero) at angles around joints. It
looks as though a number of kinesthetic and proprioceptive "signals," which
in the HPCT model we assume represent different levels in a hierarchy, are
directly available to the nervous system. But that does not necessarily
mean that the nervous system is monitoring its "output," since the signals
from those input functions (length, tension, angle, acceleration) are all
*inputs* which then become available for control.

That comment of mine suggests a tie between your questions, Hans, and the
thread begun by Gary Cziko during the past few days, concerning teaching.
(Sorry, the editor I am using today makes it difficult for me to exit and
locate the exact posting time.) In replies and continuing remarks , Bill
Powers, Hugh Petrie, and Gary have discussed the idea that a teacher helps a
learner discover perceptions that were there and unnoticed, or that were not
there at all. In either case, the teacher intends to let the learner
experience the perceptions and then use them as reference perceptions under
her or his (the learner's) control. That is the position we are all in,
at birth, and in any new learning situation -- discovering *other*
perceptions that we might come to control. We seem to start with the
kinesthetic and proprioceptive control systems working through several of
the levels for which we have direct input functions and perceptual signals,
then we start to discover a lot of other (seemingly interesting) unintended
side effects of playing with and using those systems. Out of that play
emerges our intraction with, our control of perceptions of, and our eventual
conceptualization of, a world beyond our own muscles and joints. In this
case, experience at achieving control seems to be the teacher; formal
teaching looks like a social extension of that process. I think that, at
least for people, either randon processes, or non-random processes, or both,
can be at work.

I envy all of you who can devote so much time to these fascinating dis-
cussions!

Who has time? Some of us live dangerously close to being discovered by our
heads of department and our deans. If that ever happens, we will probably
agree with Avery Andrews that social control systems are very real, indeed!

Until later,

Tom