[From Rick Marken (960409.0940)]
Chris Cherpas (960408.1534 PT) --
I doubt if I could ever be very annoyed with you [Bill Powers] (maybe Rick,
the CSGNET analogy to Huxley -- in the role of Darwin's "bulldog").
I'm honored.
Since I've done EAB work, and have now been exposed to PCT, I intuitively
read something like, "behavior Y is reinforced by some property of X" as
something like "behavior Y probably controls some property of X."
Excellent!
Both kinds of statements fit a general view that behavior is an evolutionary
process.
Well, that may be true. But, fit or no fit, one statement is demonstrably
true and the other is demonstrably false. Saying that "behavior Y probably
controls some property of X" is not just another way of saying that "behavior
Y is reinforced by some property of X". It is a way of saying that behavior
not only looks purposive; it IS purposive. Consequences don't select
(reinforce) behaviors; organisms select particular consequences and act
(behave) in order to produce them.
could anyone help me with understanding if this is the case -- i.e., that
_within_ a control system, there is no variation and selection? If so,
what are the variations, what's the unit of selection, and what's
doing the selecting
I'll try.
A control system continuously varies its output to maintain a perceptual
variable in a specified (selected) state. So there is "variation" in a
control organization inasmuch as the control system varies its output, as
necessary, to produce and maintain a particular level of a perceptual
variable. There is "selection" in a control system inasmuch as the reference
input to the control system selects the level to which the perceptual
variable is to be brought and maintained.
The reorganization system is a control system that controls a perception of
some aspect of the performance of other control systems (such as average
error over time). Like any control system, the reorganization (learning)
system varies its outputs (which, for the reorganization system, have effects
on parameters of the other control systems, rather than on environmental
variables) in order to bring and maintain the controlled perceptual variable
(average error in the other control system, say) to the selected (reference)
level (like zero).
Best
T. H. Marken