CT details; Coin Game

[Martin Taylor 920311 17:30]
(Bill Powers 920311.1100)

Yes, I could have been more precise: the subject being there is an insufficient
condition. But the point of the argument is that control cannot be detected
in the absence of a prediction about what would have been disturbed in the
absence of control.

I appreciate the Coin Game. My thesis supervisor did quite extensive studies
of this kind of game, perhaps not structured exactly the same way, but very
like. He was studying perception, not control, but the issue is also of
determining by such trials the nature of prespecified relationships. I think
it was a popular experimental paradigm at the time. The effects of interactions
on the kinds of relations that were readily detected or were detected only
with difficulty was the point at issue. He came up with the notion of
integral and separable perceptual dimensions, which seem to be quite
important. I can't give any specific references, but if you want to
search for them, look for W. R. Garner in the late 50's or 60's.

Martin, I know you have a deep grasp of control principles. I expect no
less of you. But however much one knows, there are always blind spots and
misinterpretations. Control theory reveals endless depths of new
detail, and I don't know of anyone who has plumbed them completely in just
a few years. Ask some of our Old Hands when the last new understanding came
to them, and how long they'd been control theorists before that. And think
of how much remains to be developed, that nobody has answers for!

I think you credit me with more than I deserve. There are indeed many
puzzling holes, and most of my postings are either trying to expose them for
you to fill, or to fill holes I perceive in the understanding of others.
At the moment, the hole that dominates my puzzlement is the one you began
to address in this posting, of how to assert error when the relevant percepts
are extended in time. I'm not clear what it means to have a sequence error,
though it is clear what it means to have a sequence that is in error.

In this kind of slow interaction (fast by comparison with that of Abbe
Mersenne, but slow compared to sitting around a blackboard), the problem
is to determine just what is the problem. It is often clear that a
misunderstanding exists, but the attempt to clarify it often misses the
mark, and sounds like "teaching your grandmother to suck eggs." I don't
know of any solution to this problem, except to rely on the perception that
any such attempt (a) was well intentioned, and probably helps other readers
besides, and (b) can point the way to a better understanding of where
misunderstandings really lie. We have to put up with it.

Martin

PS I have just been informed that our system will be cut off from e-mail
between mid-afternoon March 12 until some time March 16. So don't anticipate
much from me in that period!

PPS Did anyone (Gary?) ever contact Jan-Olof Eklundh?

[From Bill Powers 920311.1100]

Martin Taylor (920311) --

I see nothing in your posting, whether by diagram or in the
explanations that is not crystal clear, and has been so for ages, at
least in my mind, if not in my writings.

I think that if you take your posting together with the following
paragraph from mine of 920309 11:00, you may see why I think we have a
communication problem, and perhaps will be able to resolve it:

[...] If the experimenter happened to be correct that what she did would
have disturbed the pattern if the subject had not been there, then there
is evidence that the subject is controlling.

My ears pricked up at this. Technically, you're right -- however, the
condition isn't that there would have been an effect if the subject hadn't
been there, but that there would have been an effect if the subject's
actions hadn't canceled it. The subject might have been controlling for
some other aspect of the environment. The critical thing is that you
disturb something that seems to be affected by the subject's action, but
the disturbance isn't counteracted.

Play the Coin Game, please. The Test is done WITH the subject there. The
actions of the subject can be perceived as affecting the environment in
many ways, and objectively has many different effects on objects,
relationships, etc. in the environment. The question is which, if any, of
these effects of the subject's actions is under control. The experimenter
devises a disturbance that will alter one of those effects. If the effect
changes -- if the subject does not change the action in a way that prevents
the change from taking place -- then that effect of the action is not under
control.

The Coin Game:

Use four coins (same or different as you please). Two people play, an
Experimenter and a Subject. The Subject places the coins on a table such
that they exemplify a pattern or condition that the subject has in mind.
The Subject privately writes down this reference pattern on a piece of
paper, and hides it. The Experimenter is to discover what the controlled
pattern is, by means of disturbing the arrangement of the coins.

The rules are as follows. One round of the game starts with the
Experimenter doing something that alters the arrangement of coins on the
table. The Subject looks at the new arrangement, and if the target pattern
can still be seen, says "No error." If the pattern now differs from the
target pattern, the Subject makes any rearrangement of the coins required
so that the percieved pattern once again matches the target pattern. After
either a "No error" response or a corrective move, it is the Experimenter's
turn again.

The game ends when the experimenter can demonstrate three different moves
predicted to produce a "no error" response, and three different moves
predicted to produce a correction. Then the subject displays the written
description of the reference condition. No verbal communication except the
words "no error" takes place during the game.

You might think at first that it will be easy for the Experimenter to
discover the pattern, and compensate by choosing (as Subject) a complex
reference condition. I advise choosing a simple reference condition if you
want the game to finish in under half an hour, or not be abandoned.

This game illustrates all the facets of the test for the controlled
variable. Clark McPhail has been using it to teach The Test (he sent me
copies of the experimental reports by about 50 of his students -- wonderful
reading, especially the comment by one student that he really admired
sociologists for being able to use The Test in their work, because it is so
complex).

Martin, I know you have a deep grasp of control principles. I expect no
less of you. But however much one knows, there are always blind spots and
misinterpretations. Control theory reveals endless depths of new
detail, and I don't know of anyone who has plumbed them completely in just
a few years. Ask some of our Old Hands when the last new understanding came
to them, and how long they'd been control theorists before that. And think
of how much remains to be developed, that nobody has answers for!

···

----------------------------------
Zeros and sequences:

Suppose you want to produce a sequence like "now is the time for all good
men to come to the aid of their country." By the wasteful pandemonium
postulate, this is the province of just one sequence-recognizer and control
system.

I want to produce this effect: the sequence is recognized while it
is occurring, maximum recognition resulting when the sequence proceeds
exactly as the system is designed to recognize it, to the end. The
reference signal from a higher system doesn't contain this sequence (if it
did, the higher system would be the sequence level!), but just says "make
your sequence appear in perception." The perceptual signal isn't the
sequence, but simply an indication of the degree to which the sequence in
question is occurring. The sequence is physically present in the sequence
of reference signals and subsequent perceptual signals in individual lower-
level systems that supply the elements. So we recognize the sequence, and
can correct errors when they occur, and can produce the outputs that keep
the sequence going. It's hard to imagine how to produce the correct
sequence of outputs without defining the sequence twice: once in the
recognizer, and again in the output function.

Clearly, this system needs to be able to generate a perceptual signal
indicating that some particular sequence is in progress so far. It also
needs to be able to switch lower-level reference signals to produce the
elements of that sequence, which are lower-level perceptual signals.
Disturbances of the sequence (incorrect elements) should produce an error
and a change in the output that does something appropriate -- resets the
whole system, replaces the wrong element, and so on. I don't know how to
design a system that will behave exactly like this. I think it could be
done, but doing it with the sequence defined in only one place would be
tricky. The design will probably not look like three boxes.

This doesn't really answer your questions about zeroing, but it does
answer one question: does Bill have a design for a sequence-controlling (as
opposed to "emitting") system? The answer is no. There's some principle
missing here. These higher-level systems are hard and fuzzy at the same
time.
----------------------------------------------------------------------

Best

Bill P.

Martin

PPS Did anyone (Gary?) ever contact Jan-Olof Eklundh?

Yes, I sent him introductory information on PCT and CSGnet. But whether he
received it or not, I do not know.--Gary

···

------------------------------------------------------------------
Gary A. Cziko Telephone: (217) 333-4382
Educational Psychology FAX: (217) 244-40538
University of Illinois Internet: g-cziko@uiuc.edu
1310 S. Sixth Street Radio: N9MJZ
210 Education Building
Champaign, Illinois 61820-6990
USA
------------------------------------------------------------------