[From Bill Powers (2004.08.08.0911 MDT)]
I guess I put the wrong date on the last post.
I don’t intend to reopen any wounds, but the problem of “when are
you controlling?” reminds me of the question of “when are you
coercing?” Bjorn S. is not the only one who has assumed that if the
output of a control system is zero, the control system is not
controlling. I’m beginning to see that the problem here is that of
whether we mean an action or a relationship.
If you think of controlling as an action, and an action as a change in
the state of an output variable, then no action means no controlling. But
if you think of controlling as what a control system does, and control as
a relationship of a system to its environment, then the amount or
direction or change of the action becomes irrelevant. What then matters
is how action is related to disturbances.
As long as the control system has a power supply and is physically
intact, and as long as none of its connections to the environment has
been broken, it is controlling. If the disturbance happens to be zero,
and the reference level of the input quantity happens to be zero, then
for the moment the output of the control system will be zero. But at any
time, with no change in the control system’s organization at all, any
disturbance will immediately result in an opposing action. This proves
that the control system is “alive” – that it is turned on and
operating, just as you can prove that a silent radio is working by
turning the volume control up and hearing sound as a result.
To show that the control system is not controlling, you would
apply a disturbance to the controlled quantity and show that the output
of the system did not change, or did not change in such a way as to
oppose the disturbance.
This is exactly the problem we had with the subject of coercion. Am I
coercing you if you happen to want to do what I insist that you do?
Several people answered no, and I objected, saying yes. But I can now see
that the disagreement rests entirely on whether the term coercion is
meant as a dynamic action, or as a relationship of the coercer to the
environment. It boils down to the same definition of control: is the
coercer controlling another person if that person wants to perform the
action that the controller wants to perceive?
My answer would be to apply a disturbance and see whether the coercer
resists it. If I caused the other person to behave in a manner different
from what the supposed coercer wants to see, what does the coercer do? If
the coercer does nothing, then the coercer is not controlling the other
person, and probably never was doing so. But if the coercer immediately
produces some sort of action tending to restore the other person’s
behavior back to the pattern the coercer wants to see, then the coercer
is controlling the behavior of the other person. It’s exactly the same
criterion I would apply to determine whether any system is a working
control system or is inert or just observing.
There is a difference between saying that one person is observing the
behavior of another person, and saying that one person is controlling the
behavior of another person. And there is a difference between saying that
a control system is not acting on a controlled quantity because there is
no disturbance and thus no error, and that it is unable to act on the
controlled quantity because it has no way of doing so. When it fails to
act because there is no error, it is still activated and controlling;
there is simply no error to correct. But when there is no ability to act,
it doesn’t matter whether the error is zero or nonzero. Those two cases
coincide when the action is zero, and only the application of a
disturbance can reveal whether the system is controlling or
not.
Best,
Bill P.