Apologies all round; reorganization

[Martin Taylor 940926 13:40]

Tom Bourbon [940923.1205]

Martin:
. . .

Jeff--NO data will be accepted as valid in this discussion group unless it
was collected by a PCT-accredited individual using The Test.

As for the malarky about "PCT-accredited individuals," I find it malordorous.

You are fully justified in your comment, and I would withdraw it if I could
go back to last Friday. There's no excuse for taking out on one group of
people a frustration due to completely independent events. (Although
PCT provides a direct explanation of why one sometimes does so).

Last Friday was not one of the best of days, and I'm sorry that the evidence
of that became so clear, as an "irrelevant side effect of [failure to]
control."

ยทยทยท

-------------------

Something on which I do have a problem more worth following up is the
question of the statistical nature of evidence. My problem is not what
Tom thinks.

The theory says that reorganization is random and settles on whatever
hierarchy provides adequate control of the perceptions of the individual.
The theorists then say that data are inadequate unless they apply identically
on EVERY occasion to EVERY individual.

Martin, I believe you confuse the ideas (a) that the CRITERIA for good data
must be met on every occasion and (b) that the MODEL must apply to the
behavioral data of EVERY individual equally well every time, with the idea
that the behavioral DATA must be identical on every occasion for every
individual.

This is somewhat ambiguous as written, but I assume you mean that you
expect (a) and (b) to hold, whereas I confuse that joint truth with a
different concept--that the data should always be the same. I don't
think I make that confusion. My problem is that, given the nature of
a reorganized hierarchy, even when (b) is true the data will not allow
you to determine that fact when application of the model requires you to
consider an extended network of control systems.

When the data can be considered as having come from a single level of
control system, even from many operating in parallel, the question of
the randomness of reorganization doen't enter the picture. Control is
control, and the kinds of questions asked in the model fit are to do
with the parameters of the effective orthogonal control loops (not of
the actual control loops that may exist within the organism under study,
for these are probably distributed in ways that cannot be distinguished
by external data). But when there are NECESSARILY multiple levels of
control in the model, then the vagaries of reorganization may well determine
that different individuals behave characteristically differently, and
not differently because of overt parameter differences within individual
effective elementary control units.

The latter case is clearly impossible; the former two are much
to be desired. Reorganization probably will not make all individuals equal,
but the model (which includes reorganization) must apply to all individuals
no matter how they reorganize, so long as when reorganization ends they are
still perceptual control systems.

Yes, my thought is not that the model won't apply, but that it will, in
ways that differ among individuals.

Sorry again for the unwarranted insult. I can't promise that it won't
happen again, but I hope it is a long time before it does.

Martin

Tom Bourbon [940927.1243]

[Martin Taylor 940926 13:40]

Tom Bourbon [940923.1205]

Martin replied to my protest of his earlier remarks about PCT-accredited
individuals," The Test, and this net.

Martin's reply:

You are fully justified in your comment, and I would withdraw it if I could
go back to last Friday. There's no excuse for taking out on one group of
people a frustration due to completely independent events. (Although
PCT provides a direct explanation of why one sometimes does so).

Last Friday was not one of the best of days, and I'm sorry that the evidence
of that became so clear, as an "irrelevant side effect of [failure to]
control."

Ah, one of _those_ days, Martin. That explains the uncharacteristic
remarks. Understood, and no apologies are needed.

-------------------

Something on which I do have a problem more worth following up is the
question of the statistical nature of evidence. My problem is not what
Tom thinks.

Martin:

The theory says that reorganization is random and settles on whatever
hierarchy provides adequate control of the perceptions of the individual.
The theorists then say that data are inadequate unless they apply identically
on EVERY occasion to EVERY individual.

Tom:

Martin, I believe you confuse the ideas (a) that the CRITERIA for good data
must be met on every occasion and (b) that the MODEL must apply to the
behavioral data of EVERY individual equally well every time, with the idea
that the behavioral DATA must be identical on every occasion for every
individual.

Martin:

. . . I assume you mean that you
expect (a) and (b) to hold, whereas I confuse that joint truth with a
different concept--that the data should always be the same.

You saw through my ambiguity.

I don't
think I make that confusion. My problem is that, given the nature of
a reorganized hierarchy, even when (b) is true the data will not allow
you to determine that fact when application of the model requires you to
consider an extended network of control systems.

In other words, were we to establish n hierarchical PCT models, all of which
achieved similarly precise control of a particular perceptual signal, p (the
same "signal" in all models), and all capable of reorganizing, and were we
then to expose all of the models to the same "environmental pressures" that
were conducive to reorganization, then even if all of the models reorganized
and maintained control of p, we would not know, from that fact alone, the
new organization of any particular model. Am I close?

When the data can be considered as having come from a single level of
control system, even from many operating in parallel, the question of
the randomness of reorganization doen't enter the picture. Control is
control, and the kinds of questions asked in the model fit are to do
with the parameters of the effective orthogonal control loops (not of
the actual control loops that may exist within the organism under study,
for these are probably distributed in ways that cannot be distinguished
by external data).

Yes. I believe you are making an important point. That is exactly the case
when we fit PCT models to tracking data from several people. The results of
the modeling are consistent with our idea that all creatures that control
can be modeled as perceptual control systems, but the results would not
support strong claims about any specific person's hierarchical organization.

A couple of cases in point from some of my current work with patients. Some
of the head-injured patients (all of whom score very low on conventional
neuropsychological assessments) perform disturbed pursuit tracking as well
as I. (No jumping to conclusions, gang!) Their _results_ cannot be
distinguished from mine, but if you watch us run the tasks you will usually
see obvious differences in _how_ we do them and in many of our unintended
side effects. I believe the patients have reorganized to produce similar
results with dramatically altered "equipment." My beliefs aside, the same
single-loop PCT model can represent all of us.

Another case. I doubt that anyone could distinguish the results when our
Departmental Administrative Assistant, who is quadriplegic, or I run the
tracking tasks with a control stick strapped to the forerarm. But there
are differences. I can hold my arm in practically any position and control
equally well, but he must rest his elbow on the arm of the chair, where he
uses gravity to replace his non-functional triceps. If we support his arm
so it is horizontal, he can flex and relax at the elbow; he cannot
extend the elbow and his tracking deteriorates. His arms are pure, single
pull-only output devices and he has reorganized to produce many of the same
results as me with my double pull-only (in combination, pull-pull) devices.
The same single-loop PCT model can represent both of us.

But when there are NECESSARILY multiple levels of
control in the model, then the vagaries of reorganization may well determine
that different individuals behave characteristically differently, and
not differently because of overt parameter differences within individual
effective elementary control units.

Yes. See above.

. . . Reorganization probably will not make all individuals equal,
but the model (which includes reorganization) must apply to all individuals
no matter how they reorganize, so long as when reorganization ends they are
still perceptual control systems.

Yes, my thought is not that the model won't apply, but that it will, in
ways that differ among individuals.

We share that thought.

Later,

Tom