[From: Bruce Nevin (Mon 92019 12:37:25)]
I'll pick up on this one as a test of comprehension. Please disclose
Low correlation is not a diagnostic for control. As Martin says, there
is low correlation between too many things in the universe. Things that
exhibit low correlation with one another cannot be relevant for
scientific understanding. That includes particularly the relation
between observed outcome, "stimuli" and "responses" (actions).
It seems to me that Bill's point is rather like the old Sufi teaching
story of the man looking for his keys under the street lamp. I think
probably we all know it. A friend
stops to help. After a while he gives up.
Are you sure you lost them here?
No, I lost them over there.
Then why on earth are you looking here?
There's more light here. It's all dark over there!
The correlation that matters takes us into a place where there is less
light--the correlation of an outcome with a purpose. The outcome is
deep in murky shadow because it comprises selected aspects of a
situation, leaving out other aspects that are irrelevant. What is the
criterion of relevance? The purpose is completely in darkness because
it is a memory of perceptions that are present in the situation. This
memory is internally maintained in the "subject" of the experiment, not
accessible to the experimentor. Nothing "there" in the experimental
situation to observe!
But when you observe a series of different disturbances being corrected
you get some light on the purpose or goal. In the changing situation
something emerges as invariant. (It's more or less nearly invariant,
depending on the gain, and the invariant might itself involve a
perception of change, sequence, etc., so the term "invariance" can be
a stumbling block.)
This invariant, the observable outcome, is not directly observable as a
first-order observation. It is a second-order observation, observable
only in context of certain expectations derived from hierarchical
perceptual control theory.
But even this observation of invariance in a shifting situation of the
"subject" and its environment is not the thing that correlates with
remembered perceptions within the "subject." It *reflects* the
internally-maintained goal. Only some aspects are relevant, and those
relevant aspects might not even be noticed from the point of view of the
observer. The investigator must shift perspective to recognize what
perceptions of the observed outcome are available to the "subject."
When you have some idea what the controlled outcome is you can apply the
Test. But the Test doesn't make much sense until you have identified an
invariant in the situation from the point of view of the controlling
So the unfamiliar steps include isolating an "invariant" outcome in the
organism/environement system, selecting those aspects of the outcome
that are relevant from the point of view of the organism, and applying
the Test to verify that your guess as to what is being controlled is
This is all complicated if the organism starts controlling for something
else for whatever reason.
Low correlation is relevant only for folks looking around the lamppost.
It's a way of telling them that the keys aren't there.