interactionism

[Human_Factors_Lab]

[From Rick Marken (920818.1300)]

penni sibun (920818.1200) says:

that's what interactionist approaches try to get at: an
organism doesn't have to posit its goals _de novo_ and figure out how
the satisfy them. there's already a lot of stuff around that
facilitate doing what needs to get done. as agre puts it, we ``lean
on the world'': our roads and cars are culture are designed to make
driving down the road a plausibly easy thing to do.

I think the we (PCTers) and the interactionists may just have a
fundementally different notion of what it means to understand
something. I don't get any feeling of greater understanding about how
behavior works from the suggestion that there is already a lot of stuff
around that facilitates doing what needs to be done. The way I parse it,
it sounds like you are saying that the environment (or our perception
thereof) -- which is the stuff around-- guides (facilitates) behavior
(what needs to be done). This sounds like a verbal version of the sr
explanation of behavior that you clain interactionism is not. Sayingthat
we "lean on the world" doesn't help; suppose I want to (need to) point
straight forward? I can't lean on gravity to do that; if I lean on gravity
I'll generally end up pointing down. I really just don't get this
interacting business. As Mary Powers pointed out, interacting suggests
that the actor and environment are cooperating to produce what needs
to get done. But the enviroment doesn't care whether what needs to get
done, gets done or not; the environment is just there, doing its own
thing (I think; at least that's the model of the environment that we get
from physics -- and it works extremely well). The road doesn't care
whether you stay on it or not. Whether the road makes it easy to drive
or not depends on what YOU need to get done; some of these nice,
smooth roads may make it hard to give the kids a fun bump in the back
seat. Whether the environment is easy or not makes sense only inb
terms of one's aims in that environment; I think.

maybe someone can explain this to me. when you say ``signal'' or
``variable'' or ``percept'', it has connotations to me of a unified
thing, like a tone, or a light intensity. but when i look at the
road, i am not perceiving something like a tone. if you can explain
how all the stuff my eyes take in can be a single unified thing, maybe
i won't find it so oversimplified.

I think this is an important point. I think a number of people have a
problem with this. I take it for granted that perception is just what
afferent neural firing looks like when you are a brain (which we are). I
imagine that every different perception we have is the firing rate of a
single neuron. My mental model of this is the receptive field. We know
that certain neurons (in the lateral geniculate nucleus, for example) are
"looking" for particular patterns of light on the retina. For example, the
firing rate of a particular neuron might increase as a line on the retina is
rotated from vertical to horizontal. So this neuron is a horizontal line
detector. The faster this neuron fires, the more the perception it is
having is like a horizontal line. The orientation of the line on the retina is
the input variable, the rate of firing of the neuron is the perceptual
variable. Subjectively, I imagine that the change in firing rate is
experienced as a change in orientation. The rate of neural firing, by the
way, is the "perceptual signal" and we also call it the perceptual variable
(because it varies). Obviously, I could control this variable if I could
influence the orientation on the retina of the cause of the firing in this
neuron; and this is, indeed, how neural control systems work. As I
rotate my head (or an obhect in the world) I can bring the firing rate of
the perceptual neuron (the perecptual signal) to the reference level that I
specify. Note that the reference level is also a neural firing rate; all I'm
doing is telling (with the reference rate) at which rate the perceptual
neuron should fire. The consequence of bringing the perecptual neural
rate to the reference heural rate is to create some objective state of affairs
-- such as orienting the horizontal of the computer screen with the
horizontal line connecting my eyes.

I imagine that there are many (millions?) of neurons that detect all kinds
of different properties of the world -- simultaneously. How you wire up
a neuron to sensors so that it fires in proportion to, say, the degree of
honesty in a relationship, we don't know. But we do imagine that
something like this must be what is done by our own brain. I imagine
that when I perceive a person as honest, that very abstract, temporally
and spatially defined percept is computed by the neural nets in my brain
and results in a level of firing in some neuron (the honesty detector?);
and that level of firing is what I experience as the perception of honesty.

This is the part of the model that gets complicated (and interesting) -- for
a nicer development see Powers "Behavior: The control of perception".

Best regards

Rick

···

**************************************************************

Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)

[_Martin_Taylor]

[Martin Taylor 920819 14:15]
(Penni Sibun 920818.1200) (sorry about the [y] earlier)

  [Martin Taylor 920818 11:50]
  (Penn[i] Sibun 920817)

  > there's just tons of literature on
  >perception that suggest that perc. is a *lot* of work; in fact it
  >involves action--perhaps to the degree that p. and a. are inseparable.
  >(``active vision'' is a current buzzphrase.)

  This was cast as a criticism of PCT, but in fact it seems to be a reasonable
  description of how PCT says perception works. An ECS that is trying to
  achieve a percept can go only by its error signal.

maybe someone can explain this to me. when you say ``signal'' or
``variable'' or ``percept'', it has connotations to me of a unified
thing, like a tone, or a light intensity. but when i look at the
road, i am not perceiving something like a tone. if you can explain
how all the stuff my eyes take in can be a single unified thing, maybe
i won't find it so oversimplified.

(Aside: as a student of language, you won't be surprised to know that I find
long postings without sentence capitalization hard to read.)

Maybe the subsequent postings have cleared this up for you, but in case not,
here's another attempt.

I don't know how long you have been following CSG-L, but it seems that there
is a foundational notion missing. PCT does not deal with the human as a
single control system that controls "all the stuff my eyes take in" as "a
single unified thing." Instead, there is a vast hierarchy of elementary
control systems that make up a hierarchic control system. Each of the
elementary control systems (ECSs) do control one unified thing, but that
thing (percept) is a function of many inputs. At the very lowest level,
the inputs are sensory, but above that they include the results of lower-
level perceptual functions in lower-level ECSs. Your perception of being
in the middle of the road may be a "unified thing" but it implies the
perception of myriads of other variables, such as the intensity of light off
that patch of snow to the left, the orientation of the line projected by
the centre stripe of the road, the location of the dark corresponding to
the windshield pillar, the pressure on your buttocks... You may not be
conscious of these other percepts--in PCT the notion of percept is quite
distinct from that of consciousness--but without them your perception of
being *so far* to the left of the centre line will be less precise than it
might otherwise be.

When we say "signal" we usually mean the value of the unified thing that
has been derived from all the lower-level complex, or alternatively the
reference level with which the percept is compared. "Variable" can be
anywhere. I tend to use "CEV" for "Complex Envirnmental Variable" to
denote some structure in time and space that corresponds to a percept. One
can never be sure that a CEV actually exists "out there" but if our actions
control a percept that we feel to be based on a particular CEV, then we have
some evidence that the CEV is a useful perceptual construct. It is "real"
for us (though perhaps not for anyone else). Internally, percepts, reference
signals, errors, output, etc. are all "variables," and there are probably
many more. "Democracy" could be a perceptual function that has a particular
level in specific curcumstances (very low level these four days in the USA,
I'm afraid, for me), but if there is a CEV corresponding to my (unitary)
percept "democracy," I'm pretty sure it won't be exactly the same CEV that
affects your perception of "democracy."

Another word to be careful of in PCT terms is "action." Actions are observable
but hard to interpret. An observer sees hands and arms moving, but if those
actions were duplicated on another occasion, it is highly unlikely they would
have any useful effect. Actions affect CEVs, and the effects lead to changes
in the percepts associated with the affected CEVs. But unless the observer
knows what the CEV *is*, the actions are meaningless. And to know what the
CEV is, is to know the intentions of the actor.

You are talking, for example, with a few other people, and one gets up and
opens the door. You see the arm and hand manipulating the doorknob. But the
real question you can ask him or her is "Are you leaving, or are you leaving?"
Your two alternatives are opposites, couched in exactly the same words. You
want to know whether he or she is finished with the discussion or is coming
back after a short but necessary break. You need to know what CEV is being
acted upon. Is it one related to the interest value of the discussion, or
is it related to internal physiological states? The actions are the same,
but the implications are very different. And so it is with your question.
Two almost identical actions (word strings with intonation) have meanings
associated with the CEVs you see as possibly related to controlled percepts
in the person opening the door. All unitary, all complex in construction.

Anyway, the root misunderstanding expressed in your quoted paragraph seems to
be based on the idea that there is a single unitary percept that we control,
whereas in PCT terms, each ECS controls a single unitary percept, but there
are myriads of them involved in even very simple real-world activities such
as driving.

Hope this helps

Martin

[Human_Factors_Lab]

[From Rick Marken (920819.1400)]

Martin Taylor (920819 14:15) to penni sibun

PCT does not deal with the human as a
single control system that controls "all the stuff my eyes take in" as "a
single unified thing."

Nice post Martin. I think you honed in on a common misconception about
PCT. If I didn't already understand your point I know that your nice
discussion would have helped me towards that understanding.

Dennis Delprato (920819) says:

On the conundrum of locus of control: It is my understanding that
HPCT takes the traditional question of locus of control
...as poorly put. Rather one part of the control
does not control any other part. The entire system IS a control
system.

Well, I'd be inclined to say that control theory does place the locus
of control squarely inside the control system (not in the environment).
Variables in the environment are controlled (maintained in reference states
against disturbance) and this occurs because of the nature of the control
system itself (closed loop, negative feedback and, especially, HIGH GAIN);
not because of the nature of the environment. So whoever said that the locus
of control was inside the organism (like the cognitivists, maybe) was right;
PCT just shows why this is the case and what it imples about the nature of
behavior (it is controlled perceptual experience; I like to say it means
that behavior is a subjective, not an objective phenomenon -- the cognitivists
never got THAT part).

there is behavior and there is behavior. In
the most elementary sense, it is crucial to distinguish between behavior-1,
or physical behavior (the behavior of the moon, of molecules, of stones),
and behavior-2, or psychological behavior. Most psychology up to the
present (as exhibited in behaviorism, information-processing theory)
has usually taken behavior as behavior-1.

Nice distinction. Yes, PCT is interested in behavior-2, and sees behavior-1
as just a side-effect of behavior-2, interesting to an observer, perhaps,
but usually irrelevant to the behaving system.

In the absence of certain strategies (perhaps "the test"),
an observer cannot except speculatively identify what the psychological
event here is. [In the mind reading program]

Correct. And the main reason is because the disturbances (which are
continuously influencing all 5 "behavior-1" behaviors are invisible.
So it is truly impossible to tell (just by looking at the behavior of
all 5 objects -- or, I might add, by doing any kind of analysis of their
behavior-1 behavior) which is being moved intentionally. In fact, the idea
that ANY one is moving intentinoally would not even occur to most observers;
all the objects just move around in an arbitrary pattern. The fact that
one of the visible behavior-1's is the result, in part, of a person's
active efforts to resist a disturbance is completely invisible -- because
you can't see the disturbances. The situation is similar to the one we are
in when we just watch people behave; all that behavior looks like behavior-1
because the disturbances are invisible so the resistance to disturbance
is invisible. Control theory suggests that some of the events we see as
behavior-1 may be the result of behavior-2 -- ie. control. We can tell
which behavior-1's are the result of behavior-2 by disturbing a behavior-1
variable and looking for lack of (or considerably reduced) effect. The
mind reading program does this because it can "see" the disturbances that
the observer of the behavior of the objects cannot see; so the program
can determine which disturbance is being systematically opposed.

Best regards

Rick

···

**************************************************************

Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)