Modeling/Testing

[From Rick Marken (920329 18:40)]
a
b
c
Doesn't anyone want to comment on my "Behavioral Illusion"
post? Doesn't that strike some of you as a remarkable
discovery of Bill's? If there is strong negative feedback then
stimulus response laws are environmental laws-- not organism laws.
I would be especially interested in hearing from behavioral
scientists.

···

------

Greg Williams (920329) says to Bill Powers:

There ARE a lot of cases where g(r) is demonstrably nonzero. But it is quite a
leap to your general claim. I wish you could see that the worth of HCPT
modelling does not hang on the truth of that generality. In fact, I think
leaving the question of the truth of your generalization open, at least for
now, would boost the credibility of HPCT ideas for many behavioral scientists.
It is a matter of admitting humility until the data are in. That humility in
no way compromises the significance of HPCT in explaining organismic CONTROL,
when it is actually found to be present.

I agree that HPCT is valuable even if the general claim
(that g(r) is always non-zero) is not true. But I don't think
that claim is what affects the credibility of HPCT. It is, rather,
the implicit assumption on the part of most researchers that g(r)
is generally zero -- and that HPCT is unnecessary. What these
researchers refuse to accept is the possibility that g(r)
MIGHT be large and negative. Because of this, they see
no need to test for controlled variables. They simply
take their observations at face value and proceed to develop
models that are based on the assumption that they are observing
the behavior of a g(r) = zero system (a Z system). All PCTers
are saying is that g(r) MIGHT be non-zero and, if it is, then
they are up shit's creek without a paddle. But they ignore this.
I don't think PCT people are making grandiose claims; we
are just being ignored. I am perfectly willing to accept the
fact that g(r) is always zero. I would be surprised by such a result but
I don't want to prejudge. I say "just test for controlled variables";
that's all I ask of the people who are developing the "open loop"
models. If it turns out that the output variables that they are
dealing with are, indeed, generated "open loop" then their models
are just fine and I'll be happy; at least they eliminated
the possibility that they are dealing with a negative feedback
control loop by doing the test.

Perhaps this message has not been made clearly to the "other side".
The message is this: "We (PCTers) have no theoretical ax to grind.
We have a theory that explains control and we have used it successfully
to model behavior where control is demonstrably involved. We know
that if you are dealing with behavior that is part of a negative
feedback loop then observed input-output relationships are
mirrors of environmental (not organismic) functions. So we
heartily suggest that you test for control (negative feedback)
before you start developing models of what appears to be
cause-effect processes."

That's ALL we ask. Just test for control -- please! Without
such testing we have absolutely no way to evaluate the
success of any modeling efforts. We don't say cause-
effect models are wrong -- though we suspect that they are
most unlikely to be correct. But that is an empirical, not
a theoretical question. If all these cause-effect modelers
would just start doing the test for controlled variables, then
I, for one, would be happy -- whatever the results.

Regards

Rick

     **************************************************************

Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
Internet:marken@aerospace.aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)