Phenomenal phenomena

[Marken_Richard_S1]

[From Rick Marken (940602.1400)]

Martin Taylor (940602 14:00) --

Simulations provide proof of principle, and surprises in simulations lead to
better understanding of principles.

This is certainly true. But the main reason I run simulations (models) is to
understand phenomena, not principles. The phenomenon I want to understand is
purposive behavior.

Consider Rick's discussion some weeks ago of the experimentation on the
Marken effect. He said, if I recall, that it originally didn't work in
simulation, until they added a transport lag. He then said that the Marken
effect was explained by transport lag.

Close, but you skipped over the best part -- the phenomenon. The Marken
effect IS a phenomenon; it "works" because it happens. It is a real
phenomenon. It was discovered while I was testing what I thought was a very
obvious prediction of the PCT model -- that the response to a distrubing
variable should be the same whether this variable was the output of another
control system or of an inanimate signal source. What I found was quite
surprisingly different: the response to the other-control-system-generated
distrubance differed from the response to the SAME waveform from a signal
generator. This observation was a BIG surprise and I didn't report it until I
was sure of what I observed; my report was then quickly confired by Bill
Powers; he duplicated the PHENOMENON. This phenomenon (dubbed the Marken
effect -- my kids are still proud)) seemed to be inconsistent with the basic
PCT model. The question now was "why is this happening? How can we explain
what we see? Bill Powers quickly discovered the simple explanation -- this IS
the way a control system behaves when there is a transport lag (and the
human nervous system obviously does introduce a transport lag). So there was
no need for a new model. The "surprising" observation was surprising only
because I had not included a necessary part of the model (the need for the
transport lag in our models is not always evident when we are working with
other data).

So when you say that inclusion of the transport lag "isn't an explanation; it
is a description of the circumstances under which the effect is found" you
are, I think, missing the point -- big time. The transport lag is part of
the model that explains the phenomenon (better control with live than with
equivalent passive disturbance). Saying that inclusion of the transport lag
is a "description of the circumstances under which the effect is found" is
like saying that "having a perceptual system is a description of the
circumstances under which the phenomenon of control is found". It's a
true statement -- but it doesn't really capture what we are doing when we
are doing modelling, which is, trying to find a mechanism that will explain
the phenomena we observe.

PCT should be useful as well as interesting.

Yes. And it's use is to explain purposive behavior, NOT to say old things
in new ways or to justify what was already thought to be true.

Best

Rick

[_Martin_Taylor]

[Martin Taylor 940602 17:30]

Rick Marken (940602.0850) and (940602.1400)

I echo Bill P's compliments on your explanation of program-level perception.

My notions of that level have always been vague and you have made them
less so. But there is still vagueness in my mind, and it is captured in
your:

Of course, I could have set things up so that I had to generate something
more like a program of outputs (as Paul Revere did). For example, I could
have set it up so that, in order to control the program. I had to press an
odd numbered key when I saw a number >5 on the left and an even one
otherwise (this is like the "one if by land, two if by sea" situation where,
given the lights I see, I ride yelling the appropriate warning) -- but,
again, even in this situation, I am controlling the lower order
perceptual variables (pressing the keys) in order to maintain a
PERCEPTION of a program in progress.

I can see clearly that you are controlling the lower order perceptual
variables to maintain a perception of the program perception. But there
still seems to me to be a flavour of choosing among outputs, that is not
there when different perceptual conditions (as compared to their reference
levels) lead to different outputs in the lower levels. At the program
level, the different outputs can be qualitatively quite different, as
in "scratch your nose if the light is red, otherwise yell 'how-de-doo'".
The conditional SEEMS (and I recognize this may be more apparent than real)
to determine output rather than simply to be a controlled perception.

So when you say that inclusion of the transport lag "isn't an explanation; it
is a description of the circumstances under which the effect is found" you
are, I think, missing the point -- big time.
...
It's a
true statement -- but it doesn't really capture what we are doing when we
are doing modelling, which is, trying to find a mechanism that will explain
the phenomena we observe.

I guess my point was something like your explanation that I elided with the
three dots: that the existing control model provided the explanation, without
addition, and that the transport lag, in itself, explained nothing. It
only provided the circumstances in which the a-priori surprising effect
is manifest. Likewise, my explorations of information theory and other
physical approaches to the behaviour of control systems sometimes come
up with surprises. The theories apply everywhere, not only to control
systems, but in the control system application, surprises sometimes occur.
They are, like the Marken effect, implicit in the general case but
made manifest in the particular (control system) case.

It is interesting to watch people control for the application of IT to
subjects to which it doesn't apply.

I suppose it remains useless to point out that the question is not whether
it applies (which it necessarily does), but whether there is value in
applying it (a question whose answer is not yet agreed upon by the discutants
on CSG-L).

PCT should be useful as well as interesting.

Yes. And it's use is to explain purposive behavior, NOT to say old things
in new ways or to justify what was already thought to be true.

Is the ONLY use of a new broom to sweep new floors clean? Can't it be
used on old floors as well? I'd add "only" after your emphatic "NOT."
It's always nice when a theory has a wide scope, and can incorporate
what was already thought to be true, and can say old things, that used
to be off by themselves, in new ways that show how they are related.
After all, Euripides and Shakespear, among others, explained purposive
behaviour quite well. PCT only says those old things in new ways

(I'm torn between using the few minutes I am allotting to CSG/PCT discussion
these days for clearing up old mail and responding to new "more live"
mail. Today was the turn of new stuff. Maybe tomorrow I'll get back to
the old.

Bill P., there's far, far too much to disagree with in your mailings of
today for me to deal with it in the time available. Sorry.)

Martin