[From Rick Marken (940406.0830)]
Bill Leach (940405.22:08 EST) --
as long as the reference remains unchanged, a change in preception from ANY
cause will cause an output change, INCLUDING a change in the disturbance.
Yes. That little "ANY cause" part makes all the difference. It makes
it clear that distubance caused change in perception is not THE cause of
output change; perception changes partly due to ongoing output changes as
well.
My only problem with your otherwise correct statement is that it describes
only half of what is going on in a control loop. You are saying that
do/dt =f(dp/dt); change in output is a function of change in perception,
which is certainly true. But, it is only half the story. It is also true
that, AT THE SAME TIME, change in perception is a function of change in
output (and disturbance): dp/dt= g(do/dt) + h(dd/dt). The simultaneous,
"quasi-statuc" solution to the algebraic versions of these differential
equations shows that p = r (approximately) and that o = r -1/g(d). So in an
operating control loop, p is a "dependent variable" -- the value of p being
determined by the reference signal, r. As r varies, so goes p. The output,
o, does not depend on p at all; it depends on r and d. When r is fixed, it
depends only on d.
So, while it is true that we can analyze the loop into causal components, one
of which is the causal influence of perception on output, in the working loop
these causal links are "absorbed" into the functioning of the loop.
I think it is important to understand this because, when dealing with living
systems, it is easy to be seduced into thinking that the causal link between
perception and output (which we know exists from analysis of the loop) has
some special significance. Indeed, the way life scientists think about
behavior is based on behavioral "imagery" that views the effect of perception
on output as the one that is of central importance.
Analysis of the closed loop shows that the effect of perception on output is
no more nor less important than the effect of output on perception; there is
a causal loop. PCT is trying to replace the "perception as cause" way of
looking at behavior with a new imagery that doesn't single out one half of
the feedback loop for special attention; the PCT imagery is that perception
is controlled. Perceptual variables are causes and results at the same time;
they are controlled variables.
The reason I argue against information theory is not just because it is
irrelevant to control theory; it is because it is based on the wrong
behavioral imagery; it is a misleading way of looking at the behavior of
living control systems. Information theory (as used in psychology) is based
on the idea that the effect of perception on output -- the do/dt=f(dp/dt)
part of the loop -- has a special status; information is associated
with perception and it is used, in some way, to determine output. Information
theory focuses our attention on perception as "cause" of output ; but that's
just half the story. Perception is just as much a RESULT as it is CAUSE of
output. Now we know (from PCT) what that means; it means that perceptual
variables are CONTROLLED. So we can dispense with models of behavior that
were based on a preference for seeing behavior in terms of one half of the
control loop or another. We lose nothing by doing this and, I believe, we
gain something in the bargain -- a clarity perspective that allows us to
focus our research efforts where we should; on the study of control.
Best
Rick