Still on the importance of behavior

[millerd]

[Dan Miller (930414.1315)]

Rick Marken and Bill Powers (both 930413) took my poorly thought out
post as an opportunity for some instruction. And, boy, I could use
some, maybe a lot. That's for sure.

Rick says that his Blind Men article is widely distributed but
seldom read. Isn't that how it goes? I give this reading to my
social psychology class and I'm sure that only the better, more
interested students read it. Those that do are in total agreement,
however. We all agree that we shouldn't think about behavior as
a response, as a release, as reinforced, or as preprogrammed
instructions. This paper sure disabused us of those worn out ideas.
We understand that behavior is the control of perception.

I must say that I am still a little confused, maybe a lot confused.
My wife loses her patience because she thinks that I am too easily
and too often confused. You all might lose your patience with me
as well, so I won't take up too much valuable time and space on
the net. More than likely, my confusion is a simple misunderstanding.
I have been known to make mistakes in the past, and I may be
mistaken now. Lord knows, I'm human. Still, I don't want to be
confused, nor do I want to confuse the model (you know the one).
So, I feel that I must ask some questions.

Rick, your discussion of Fred Astaire and Ginger Rogers really
brought back the memories. I just love those old movies, and What
Dancing. When my wife and I would watch Fred and Ginger dance I
would ask, "How do they do that?" Barbara would glare at me, so
I would ask the question to myself and try to figure out how it
was done. I'll admit it, my thoughts got pretty complex, and
more than a little confused. But Rick tells me that ...
<what you are actually seeing is the observable side effects of an
<exceptionally skillful process of controlling perceptual inputs.
<The end result of the cognitive commands are not outputs; they are
<perceptions.

Now, I'm afraid that I'm more confused than ever. I'm just not
seeing it the right way, I guess. But it seems to me, and correct
me if I am wrong, that their (Fred and Ginger's) social behavior
just seems to me to be more than observable side effects. When I
study the model, (I won't draw it here. I'd probably make a mistake
anyhow.) I can't help but think that behavior (output) is important.
To me, it was the output that Fred and Ginger were perceiving, and
it was their output that was so nicely coordinated. That is, they
were fitting their behaviors together in a very highly coordinated
fashion through the process of controlling perceptions (two models
interacting like what Mary Powers wrote about). This is pretty
complex stuff for me. I can hardly understand how two people
talk let alone how they dance.

But I digress. Back to the model - a simple ECS will do. When I
think out loud to my students, "Behavior is the control of perception,"
and I look at the model on the blackboard I cannot reconcile that
behavior is only a side effect. It just seems to me that it is as
central a component of the system as perception, reference signals,
and disturbances (like the social world - Some Disturbance). In fact,
I have gone on about this confusion so long that a couple of my more
interested students presented me with a gift. They named a principal
after me - really after my open questioning and obvious confusion.
They figured that if I had my own principle that I wouldn't be as
confused. The principle - Miller's Principle - is this:

    No Behavior - No Control

I was flattered by this gift, but more than a little concerned that
this behavior wasn't conditioned, released, reinforced, or prepro-
grammed. They assured me that it wasn't, that it was infact output
that controlled input. "You know - no behavior, no nail pounding,
no house built." Students can be so direct, can't they? "No
coordinated behavior, no dance." How coordinated? Good question.
See the model and build complexity, I guess.

But, I'm still concerned that I don't understand the model. And
then there is this complexity business. It seems like almost
everything that I am interested in has this social interaction
quality to it - dancing, conversing, making love, bargaining,
negotiating, gang activities, and so on. I know that these are
only words, but aren't they all? Bill Powers tells me that we
need to get past (below?) the words to the process. I couldn't
agree more. That is what we need to do, get to the process.

Later,
Dan Miller

[_Martin_Taylor]

[Martin Taylor 930415 11:30]
(Dan Miller 930414.1315)

Back to the model - a simple ECS will do. When I
think out loud to my students, "Behavior is the control of perception,"
and I look at the model on the blackboard I cannot reconcile that
behavior is only a side effect. It just seems to me that it is as
central a component of the system as perception, reference signals,
and disturbances (like the social world - Some Disturbance).

There's a way out of this, and that is to note that the simple ECS
diagram is not a complete diagram of the interactions of the control
system with the world.

                 > >
                 > >
                 > >
          -------|------- --------|--------
         > > > > > >
         > -----O---- | | -----O---- |
not | | | | but | | | |
         > PIF OUT | | PIF OUT |
         > > > > > > > >>> >
          --^---------V-- -^ ^--------VVV--
            > > / \ / | \
             <--CEV<-- | --X<-- | \
                 > \ | | \
              disturbing \ distx | V
              variable \ / other effects
                                             --Y<-- not observed by
                                               > the sensory input
                                              disty system of the ECS

In my way of looking at PCT, which some call "an S-R view" so it can't
be authentic, the CEV (complex environmental variable) is a construct
of the PIF (perceptual input function). Its state is what is perceived
as the scalar perceptual signal. In the standard diagram, it is treated
as an entity in the world, that other observers could monitor. It is
affected by a disturbing variable that would cause a disturbance, were
it not for the fact that the CEV is affected by the actions consequent on
the output of the ECS. In the standard diagram, affecting the CEV is all
that the output does in the world.

In a diagram that I see as more correct (though still oversimplified), the
CEV cannot readily be discerned by an outside observer, since it is (as its
name implies) a "complex". In the diagram, I show an X variable and a Y
variable, affected by different kinds of disturbance. The CEV is generated
in the way that the PIF relates X and Y to form the perceptual signal.
The output affects the world in various ways; it affects X and Y in such
a way that their complex relationship is maintained in the controlled
value of the perceptual signal, but in addition, it has side effects,
influences on aspects of the world not detected by its PIF.

In my version of PCT, which I hope to be orthodox in this respect, the
effects of the output on X and Y that are part of the control loop are
"behaviour," whereas the totality of effects on the world that can be
observed by an outsider are "actions." Behaviour is not a side effect,
under this definition. It is intrinsic. Neither are actions side effects.
Actions may HAVE side effects. The ECS knows nothing about these side
effects, though later they may react upon X and Y in some manner. If
they do, they would then be considered to be part of "distx" and "disty,"
the relation with the prior output being undetected.

···

----------
An additional way out of your dilemma is to note that a single ECS does
not exist alone. Other ECSs use the same output effectors at the same
time: "one seldom controls only a single perception at any level." Actions
help satisfy many goals at once. So, any particular output has an
indeterminate relation with the environmental elements that enter into
the CEV that is determined by its PIF. What is constant is the control
of the scalar signal determined by that PIF. So, the actions involved
in controlling any specific perceptual signal may differ substantially,
even qualitatively, from moment to moment. Speaking loosely, then, one
might say that the SPECIFIC actions are side effects of control, though
at the moment they are performed, the actions are an intrinsic component
of the loop. If one is blocked, another may be enhanced. Such variable
actions LOOK LIKE side effects, though they are not.

Both approaches to the issue are appropriate, and don't conflict with each
other.

Hope this helps, and doesn't stir up the bonfires of the heretic-burners.

Martin