[From Rick Marken (941022.1800)]
Bill Powers (941021.1635 MDT), adding to the off diagonal elements
in my "berating" matrix, says:
Tom Bourbon and Rick Marken (various) --
Let's try not to totally confuse those who are trying to understand
PCT. Perceptions DO cause actions.
OK. As long as we still agree that actions are not based on
information about the distubance in the perceptual signal, I'm happy;-)
While it is certainly true that perceptions cause actions in a control
loop, I think it is important to try to keep the focus of would-be PCT
researchers/modellers on the fact that, in a functioning control loop,
perception is controlled; the causal role of perception (even with the
reference signal fixed at zero) is not visible in a control loop unless
the gain of the loop is so low that it is virtually open. So, while it's
true (assuming a fixed, zero reference signal) that o = f(-p) [action, o, is
caused by perception, p] it is also true, _at the same time_, that p =
g(o) + h(d) [ perception is caused by action and disturbance]. Because
these relationships occur in a loop, you cannot observe the actual causal
relationship between action and perception (the function f()) by
monitoring the behavior of the variables in the loop; the relationship
between the observed inputs and outputs of a closed loop system tell
you nothing about the nature of the causal connection between
perception and action. You can only get at this relationship by
modelling (guessing) and testing goodness of fit (as you do in the
measurement of the closed loop transfer function). What you do see
in a control loop is the relationship between disturbance and action and
(if you know how to do it) the relationship between controlled
perceptual variable and reference signal.
When we model control, we must guess at the causal connection
between action and perception . We also guess (if we don't know it
from The Test) at the perception that is being controlled; when we do
this we are guessing the perception to which the control system is
responding. But I think it is not a good idea to think of it this way
because, when we do, we are apt to think of a perception as the
cause of the actions _that we observe_. So the inclination is to look for
perceptual variables (actually the environmental correlates thereof)
that might cause the actions we observe. In fact, the actions we
observe depend on disturbances and the feedback function, g(), and only
indirectly on perception.
The idea that perceptions cause actions is the basis of modelling and
research in conventional psychology. It is the wrong approach, not
because perceptions don't cause actions (they do) but because they
cause these actions in a closed loop. The conventional approach ignores
half the story (the effect of actions on perceptions) and takes the
observed relationship between disturbance and action as evidence that they
have the whole story. The best way to automatically get the whole story
straight (as PCT does) is , I believe, to keep the focus on the fact that, in
a control loop, a perceptual variable is controlled. All kinds of different
actions result when the same perception is controlled in the face of
disturbances and changes to the feedback function. It's easy to
overlook this important fact about control system behavior when our focus is
on the cause of action by perception instead of the control of perception
by action.
Best
Rick