[From Rick Marken (941228.2045)]
Bruce Abbott (941228.1150 EST) --
Given that reference, the consequences determine what will happen
on the next loop of the program. This is all that I mean by "selection
by consequences"
You are doing an excellent job of trying to preserve (for yourself) the
notion that "selection by consequences" is consistent with PCT. Now
you claim that you have always assumed that there is a reference for
the state of the consequences; so consequences select (determine) action
only with respect to this reference.
I'm curious, Bruce. Is this the way behaviorists think of "selection by
consequences"? Did you always think of "selection by consequences"
this way? That is, did you always know that organisms have references
for the consequences of their actions? Did you always know that these
references were for perceptual aspects of the consequences of actions?
Did you always know that these references are varied by the organism
itself so that the same consequence can "determine" quite different
actions on different occasions? In other words, did you always know
that organisms were perceptual control systems? If so, is your research
methods text all about how to study perceptual control systems? Has
your own research been aimed at determining the perceptual variables
that organisms control and how they control them?
Inquiring minds want to know;-)
Anyway, your attempt to make the term "selection by consequences"
PCT friendly still seems to me to fall sort. Consequences only "select"
behavior (in your sense) when 1) the reference signal is a constant 2)
the error signal precisely determines output AND 3) the charateristics
of the output function are always the same. Under normal circumstances,
none of these three conditions are met in a living control system; all
must be met, always, if your notion of "selection by consequences" is
to make any sense as a description of the operation of a control system.
Your version of "selection by consequences" is represented by the
following equation:
o = r-p
Output (action), o, is determined ("selected") by the consequences, p, of
action if the reference signal, r, is constant. According to this equation,
with r constant, variations in o are strictly determined ("selected") by
variations in p. But r is not a constant in living systems; organisms are
continuously changing what they want to perceive; they must do this in order
to maintain control of higher level variables.
Strike one!
Also, in neural control systems there is "neural noise" (epsilon) so:
o = r-p+epsilon
This means that, even if r were a constant, o is not determined by
consequences alone; there can be a fairly substantial noise contribution
to output (the closed loop neatly compensates for neural noise so that it,
like external disturbances, has very little influence on the controlled
variable, p).
Strike two!
Finally, error is amplified into output; this amplification is carried out
by the output function, so
o = f(r-p+epsilon)
The output function,f(), represents things like the tranformation, via
muscle contraction, of a neural signal (r-p+epsilon) into a force output,
o. This transformation depends on, among other things, the state of
fatigue of the muscle, which is changing all the time. So the output
function is changing all the time. Even if r were constant and there
were no neural noise it would STILL not be true that consequences
determine ("select") actions.
Stee-rike three! "Selection by consequences" is out of there as a
description of the operation of a living control system.
However, in the interest of clarity in communication, I'll try to
refrain from using this phrase [selection by consequences] in the
future (except, of course, to annoy Rick). (;->
The phrase doesn't bother me as much as the prospect of seeing you fail
to grasp the basics of PCT simply because you won't abandon old and
beloved misconceptions about the nature of behavior. If you want to
keep saying that behavior involves "selection by consequences" it's OK
with me as long as you DO things that contribute to the development
of PCT science, such as research aimed at investigating what
conventional psychologists don't even know exists -- controlled
perceptual variables. If you do quality PCT research and modelling, like
that done by Tom Bourbon, then you can say "behavior is selected by its
consequences", "social systems are control systems", "behavior is the
control of output", "there is information about the disturbance in
perception", or anything else that would otherwise put me into orbit.
I just want to see a little PCT research; is that so wrong?
Do I perceive correctly that you and Rick now agree with me that this
simulation DOES provide an analog of the reorganization process?
I think it does provide such an analog. One major difference is that the
PCT reorganizing model assumes that the variables controlled by the
reorganizing control system are continuous, not binary. There's more
to survival than surviving and not surviving. It also makes for better
control.
Best
Rick