This, That 'n The Other Thing

[From Rick Marken (930923.0900)]

Avery Andrews (930923.0943) --

Can you describe a concrete case where real problems are caused by some
manual control engineers failing to understand the difference
between the `objective' quantity that the customer of a manual control
system wants to control, and the perception that the operator actually
controls? It's my impression that professionals tend to be pretty
resistant to conceptual points until they see a concrete illustration
of how understanding the point will help them do their jobs better

I was more interested in getting the manual controllers to under-
stand the science rather than the practical implications of PCT.
I can't think of any particular problems caused by the particular
misunderstanding you mention -- nor can I think of any particular
successes that have come from improper application of control theory
in 'human engineering' I do know that all of the manual control
"aiding" systems (predictive displays, quickened displays, etc)
that exist are based on the assumption that the engineers
know the variable being controlled (which they probably do)
and the FIXED reference for that variable. I developed a control
aiding system (suggested by Bill P. based on some research of
mine on conflict) that is reference level independent. I was going
to write it up but aiding doesn't seem that important anymore
since the actions of operators of modern aircraft are no longer
used to compensate for disturbances to attitude variables -- they
are used to set the reference of lower level (electro-mechanical
control systems -- the avionics of the aircraft) systems that
control attitude relative to these operator selected references.

Some classic "human engineering' mistakes (not necessarily
made by 'manual control theorists' per se) do reflect a lack
of understanding of some basic qualitative characteristics of
control systems. Three Mile Island happened largely because
the designers of the control panels didn't understand that
people can only control perceptual representations of objective
states of affairs. The operators at TMI were given an indication
of what action had been taken -- not about the results of that
action. So the operators knew that they had pressed the button
to open a pressure release valve -- they just didn't get an
indication (perception) of the actual state of the valve. This
is actually a good example of why saying "give people feedback"
is not necessarily a very helpful human engineering principle.

Hal Pepinsky (930923) --

I'm troubled that human beings using PCT take the position that
PCT does not answer a question as about choice of referents,

I don't understand this. PCT does try to answer questions about
choice of reference signal levels (goal selection) -- we just
haven't done much research on it. But we have built models and
done some research to show that certain behaviors can quite
accruately simulated by a hierarchy of control systems, where
higher level control systems achieve their perceptual goals by
varying ("choosing") the reference signals that specify the
perceptual goals of lower level systems.

and therefore that they themselves have no interest in
exploring whether PCT could be complemented by other models
which do indeed account for what within PCT is unaccountable.

I'm very interested in exploring it; I just haven't seen it.
I think you use the word "model" as a synonym for "a way
of looking at things". PCT is a working model of systems (like
you) that can look at things in different ways and (like I)
that can build models of those systems.

PCT is no less important when it becomes part of a larger whole.

I'll buy that. I'm not controlling for PCT being more important
than some other idea; it's not a religion. PCT is important to
me because it points to and explains the phenomenon of purposeful
behavior; that covers a lot of what living systems do. If PCT
is part of some larger whole, that's fine; I would just like to
see in what sense that's true. What is the "whole" that PCT is
part of? How does PCT fit in?

Never mind what PCT does. Why do you choose to do ONLY that?

I don't understand? I'm also choose Bach, Mozart, Dylan (old),
Jesus, Darwin, Homer, Shakespaere, Newton, Joyce, Marx (all of
'em -- Groucho, Chico, Harpo and Karl) and now (after those
nice comments about my posts -- even if done with his usual
sarcasm) Cziko.

I guess I don't understand your post, Hal. But I'm really
interested. Could you try to expand on it a bit. I think
you are concerned that PCT is not sufficiently involved
in real world human problems. Is this your concern?

Bruce Nevin (Thu 930923 08:45:44 EDT)--

Rick, you have showed us that you can't tell what someone is doing by
watching what they are doing. Look at Greg's suggestion in this light.
How can you test whether Wickens is controlling a perception of
perceptual control? Wouldn't you have to engage him in some sort of
dialog in order to do that?

It's true that you can't tell "in principle" what someone is doing
just by looking. But there are some pretty strong hints that come
from looking -- you just have to test them to get an accurate
result. For example, when you see someone bend and reach toward
the pavement on a New York street you can be pretty sure that they
are either trying to pick something up or duck something -- but
they might just be stretching. To know exactly what they are doing
(controlling) you have to test; but your initial hypotheses for
"the test" comes from looking. Looking at the person bending over,
you can eliminate some hypotheses right off the bat; they are pro-
bably not controlling the distance between their head and the
moon; they are probably not playing the great G-minor fugue, etc.
Observations do constrain your initial hypotheses about what variable
is being controlled when you do the test; it's not easy to tell
what variable IS being controlled, but it's pretty easy to tell
what variables are NOT being controlled. Thus, a psychologist who
tests tons of subjects in different environmental conditions is
probably not doing "the test for the controlled variable". But if
that's one of your hypotheses, then great -- you test it.

And I'm not a big fan of "dialog" as a way to do "the test". Try
the "coin game" and see the arguments you get into about that
your subject is "really" trying to do.

These are clearly examples of the test for
controlled perceptual variables. Does this mean Harris understood
perceptual control?

If he really was doing a version of "the test" then, yes, regardless
of what he called it, he understood that people control liguistic
variables and he was testing to determine what variables were
being controlled. A person couldn't do "the test" at all without
having at least an intuitive feel for the nature of control (that
people are keeping certain results of their actions in reference
states -- maintained against disturbance). But I can't tell from
your description of the "pair test" whether it is actually a
version of "the test". Maybe if you described the pair test in more
detail I could see if it is.

You can witness
disturbances and countervailing behavioral outputs without actually being
the author of the disturbances. This is a valid (naturalistic) form of
the test.


If Wickens and others are very close, do you object to others helping
them to recognize perceptual control and its importance?

Of course not! I WANT people to "recognize perceptual control and
its importance" What I don't want is people saying "this is how
control works" when it's not. My (waking) nightmare is running into
people who tell me all about PCT -- and it turns out to be nothing
more than a re-verbalization (in control theory terms) of the SOC
(same old crap) -- SR, cognitive, drive theory, etc.

Are you
committed to a perception that it is useless to try to do so?

How does one commit to a perception??

Why don't you just give me a run down on why you think this
"pair test" is a version of the test for the controlled
variable? Then maybe I could start interfering with your
understanding of PCT.