[From Rick Marken (960904.1030)]
Me:
The gross mismatch between MCT model and human behavior is rather convincing
evidence that the MCT model >is not the appropriate model of human
controlling.
Hans:
That may be a scaling artifact. If you look more closely, you see
that the MCT controller initially controls in the wrong direction as
well. But after a short number of observations it is back to par.
I will check it again: there may be an instant when the MCT controller
initially controls in the wrong direction; I think I did see one point that
was sitting way off the curve. But, still, it's clear that the behavior of
the MCT controller, even with this momentary response in the wrong direction,
does not look anything like the behavior of the human controller.
In our experiments, polarity reversals consistently resulted in a smooth,
exponentially increasing "runaway" responset in both the human and the simple
PCT model. What happens is this: after the polarity reversal, a handle
movement (response) in the direction that would have reduced error (the
distance from the cursor to the target) now increases error. The increase in
error leads to a continuation of the response in the same direction, a
direction that now (due to the polarity reversal) futher increases the
error. The increasing error leads to a continuation of the response
that futhre increases error: error is now causing responses the increase,
rather than decrease, error.
The polarity reversal suddenly puts the human and the model into a _positive_
feedback relationship with respect to the controlled input. The human stays
in this positive feedback relationship for about 1/2 second, apparently until
another control system (in the person) sees the problem and reverses the
output polarity of the control loop that is controlling the cursor. During
the 1/2 second positive feedback period, the human's behavior matches the
behavior of the PCT control model exactly.
The 1/2 second positive feedback runaway occurs even if the human knows in
advance that the polarity reversal is going to occur -- and even if the human
has experienced the polarity reversal several times in the past. Thus, in our
version of the polarity reversal experiment there were plenty of opportuities
for people to develop a "model" of the result of a reversal. Nevertheless, the
behavior of these people is never even slightly similar to the behavior of
the MCT model. It is quite clear that the MCT model does _not_ capture the
nature of human controlling in this situation.
Does this failure of the MCT model to explain controlling in this very simple
situation have any influence on your appraisal of the merits of MCT as a
model of the purposeful behavior of living systems?
RSM
