[From Bill Powers (2003.02.27.1024 MST)]
Marc Abrams (2003,02,26) --
>I believe the "Decision", or maybe the "Determination", ( which may be a
better word.) >of what gets controlled, both consciously and unconsciously
is determined by "Anxiety" >or "Stress". (Using capitals to indicate nouns,
not adjectives) By "Anxiety" or >"Stress" I mean "anticipation" of "what"
should happen next after a Decision, or >Determination has been made by the
Control System at that level in the hierarchy. This >of course happens when
a variable is compared to the goal from the next highest level. >It becomes
CV ( Controlled Variable ) after error has been "determined" When
the >environment or imagination , affects it causing the new &/or old CV's
to become >input for a new round of control.
What you're investigating here comes very close to the thoughts I was
having in the 1950s about reorganization, or rather the so-called
reorganizing system. What is it that gives rise to the hierarchy of control
systems, the hierarchy that uses overt behavior to achieve its ends? And
why do individuals learn to control one set of variables but not a
different but equally possible set? Even more puzzling: how is it that
satisfying hunger can lead to learning how to operate a particular
unnatural piece of apparatus? The lever in a Skinner box, when properly
manipulated, causes food to appear not because of any natural law or
long-established property of the world that could lead through evolution to
such a skill, but because of the whim of a young professor of psychology
who wanted to control an animal's actions. We're talking here about
whimsical and arbitrary "laws of nature" -- yet animals are able to learn
even such nonsensical rules as walking in a figure eight to make bits of
food appear.The question on my mind was, "What kind of system could learn
what to do under circumstances as random-seeming as that?"
Clearly, this system could not _reason out_ what to do a priori. There is
no logical connection between alternating left and right turns while
walking and the withdrawal of a cover over a food dish. This kind of system
could not be inherited; until Skinner did the experiment, this situation
had never been encountered before (I would think) in pigeondom. And I am
extremely reluctant to conclude that pigeons have the same sort of logical
rationality that we have, so they can work out what is increasing the
probability of getting fed.
With regard to that last thought, it also occurred to me that we have to
account for more than the acquisition of feeding behavior. We have to
explain the growth of rationality itself, from the rudimentary form found
in neonates to the capacity for complex and abstract reasoning in the adult
human being. We have to account for the growth of perception and control at
_all_ levels. Control systems appear at a furious rate in babies -- see the
work of Frans Plooij and his wife Hedy. It was obvious to me then, as it
still is, that whatever system it is that produces new control systems, it
can't rely on the very capacities that it will some day build, such as
rationality.
This is what led to the concept of _random_ reorganization. The insight was
that if the organism could simply alter its own neural connections at
random, it would have repeated chances at producing new systems that
controlled better than the old ones. But the main part of the insight was
how the right new organization could be made to stick rather than just
being reorganized away again. All that would be needed would be to monitor
the results of the random changes, and when a better result occurred, to
_stop the reorganization_ (or change its focus). It's as if you could base
your choices of action on a coin toss, but before you tossed it you could
see the results of the previous toss. If the results of acting according to
the previous toss were favorable, you just wouldn't toss the coin
again.This is the exact opposite of the concept of reinforcement.
Then came the 64000-dollar question, back when a thousand dollars was a
thousand dollars. How can this system know whether the result of a coin
toss was favorable or unfavorable? For this sort of system to work as
envisioned, there has to be some variable (or set of variables) that (a)
indicates whether the results were better or worse, and (b) can be sensed
by the system producing the reorganizing. We are now beginning to tie this
exposition to the subject of Marcs's post.
So what variables might there be that could indicate the status of the
organism as a whole? Well, we know of many of them: blood glucose
concentration, blood acidity, level of adrenalin and consequent
vasoconstriction, heart rate, blood pressure, respiration rate, and quite a
few others -- all of which are involved in one way or another with the many
inherited homeostatic (and homeorhetic) systems of the body. Some of these
variables can be directly sensed, neural or chemically, and others can be
sensed by reliable but indirect effects on other variables. What we are
talking about here are exactly the kinds of variables mentioned in
connection with the General Adaptation Syndrome.
Only one question remains: what determines that a particular state of any
one of these basic physiological variables is to be considered abnormal,
indicative of a need to reorganize? Any PCTer should be able to supply the
answer: a reference signal and its associated comparator. For each
variable, there is an inherited reference signal that specifies the
preferred magnitude. This could be a fixed inherited biochemical signal, an
inherited built-in bias or threshold in the sensor (neural or biochemical),
or an actual signal carried to a comparator from another system, in which
case we would have to consider homeorhesis (see Myrsovsky) and the
possibility that some intrinsic reference signals can vary over time. The
effects of these reference signals (with their associated control systems)
have been known for well over a century: they produce the stability of the
"milieu interieur" as Bernard called it, or the state of the systems that
Cannon called "homeostatic".
And what, then, is stress? If we're going to retain the exact engineering
meaning, we must distinguish between the force or influence applied from
outside, and the resulting changes, if any, that are observed. The external
forces are called "stresses" (or stressors, sometimes), and the effects
they have are called "strain." So withholding of food is a stress or
stressor, and the resulting fall in blood glucose is a strain induced by
the stress. But this is not a simple cause-effect system. Before the strain
becomes dangerous, the homeostatic systems release stored glucose to
prevent any further strain. Only when the stress has been present for a
long time, and the reserves of glycogen become depleted, does the strain
begin to increase according to the size and duration of the stress. The
blood glucose now begins to drop seriously, and the difference between the
actual concentration of glucose and its reference level starts a sharp
increase. The amount of strain, of course, cannot be estimated without
having an estimate of the reference state.
That, or in that neighborhood, is the situation when random reorganization
commences. It will not cease until the organization of the
behavior-producing central nervous system has changed in such a way as to
halt or reverse the strain, the intrinsic error, the drop in blood glucose.
If the required new behavior is simply to spot little dots of a different
color and ingest them, the reorganization may not last long. If what is
required is to walk, then turn left, then turn right, and finally to turn
left and right in alternation, the process may take longer and require some
help, such as giving food for any move in the right direction (Skinner
called that shaping).
As for most behavior, we end up with two answers to the question of what
causes it. One answer is the reference signal setting. If there were no
reference signal, no preferred level of glood glucose concentration, there
would be no error, no strain, and therefore no motivation to reorganize. If
you were faced by a life-threatening surgical procedure, but didn't care if
you lived through it, there would be no strain.
The other answer, of course, is the nature of the environment on which one
must act to have an effect on the error signal. In some environments, a
hungry pigeon can obtain grain just by searching a little further outside
the usual pecking grounds. In others, the pigeon might have to peck in a
special place, or walk in a figure eight, or peck on the image of a
battleship. If there is no action that can remove or at least lessen the
strain, and the stress continues, reorganization will simply continue until
death.
This is the most basic level at which we can consider stress and strain,
for which read "disturbance and error". Similar things can happen, I
suppose, at higher levels, and it is quite probable that any real
reorganizing system will have to have added capabilities, such as the
ability to reorganize the right aspects of behavior, those that have some
effect on the error. Future workers will decide such questions.
There was a time when I heard the phrase "stress and strain" in the same
way as "cease and desist" -- as pairs of words from different roots meaning
the same thing. But stress is not the same as strain, I learned much later.
We can take advantage of the distinction here.
So, Marc, this is my expansion on the brief telephone comment. As you can
see, I do indeed see the GAS as being intimately related to reorganization,
at least when the errors get large enough to produce reorganization.
Best,
Bill