[From Bruce Gregory (2003.0630.1414)]
Rick Marken (2003.06.30.1030)]
OK. It looks like you get far more satisfaction from arrogantly
bashing the
scientific work of others than from doing any scientific work
yourself. Enjoy.
There's nothing quite so satisfying as arrogantly bashing the work of
one's betters, nicht wahr? God, how I love the smell of napalm in the
morning!
[From Rick Marken (2003.06.30.1450)]
Bruce Gregory (2003.0630.1526)--
> Rick Marken (2003.06.30.1140)
>
> Conflict occurs when two systems at the same level in the hierarchy
> (two systems controlling the same types of perceptual variables)
> are controlling non-orthogonal (mathematically equivalent) perceptual
> variables relative to different references. Conflict is a perceptual,
> not an output, phenomenon. Conflict exists when two control
> systems act to keep what is virtually the same perceptual variable
> in two different states. In order to do this, the two systems are
> _usually_ (depending on disturbances) trying to get the same output
> into two different states.I follow you up to this point.
Great.
> But this is not what defines the conflict.
What _does_ define the conflict, then?
The perceptual variables that are controlled by the systems involved in the
conflict.
> In a hierarchical system, higher level systems are almost always
> trying to get the same lower level outputs into different states
> (in the sense that the reference for the lower level systems is
>the sum of the outputs of several higher level systems).
I'm truly lost. I guess I don't understand conflict in the sense the
term is used in HPCT.
Don't give up. Let me try to make it clearer.
Suppose we have two control systems each controlling a different perceptual
variable: p1 and p2.
Let p1 = x1+x2 and p2 = x1-x2
where x1 and x2 are sensory variables. The values of the sensory variables depend
on the values of physical variables, q1 and q2: x1 = k1*q1 and x2 = k2*q2. q1 and
q2 depend on physical disturbances (d1 and d2) and system outputs (o1 and o2) so
that:
q1 = o1 + d1 and q2 = o2+d2
The system must vary o1 and o2 appropriately in order to control p1 and p2 (keep
them equal to r1 and r2 respectively). The two control systems can simultaneously
and successfully control p1 and p2 if the outputs are a function of error as
follows:
(1) o1 = e1+e2 : o2 = e1-e2
where e1 = r1-p1 (the error in the system controlling p1) and e2 = r2-p2 (the
error in the system controlling p2).
If you set this up in a spreadsheet (using iterative calculation) you will find
that the two control systems can control their perceptions without conflict; one
control system keeps p1 nearly equal to r1 and the other control system keeps p2
nearly equal to r2. They do this even though they are "trying" to get the lower
level outputs to different levels. This can be seen in equation (1). The system
controlling p1 is trying to make o1 and o2 equal e1; the system controlling p2 is
trying to make o1 equal e2 and o2 = -e2. But this produces no conflict because
the correct value for o1 to have is e1+e2 and the correct value for o2 to have is
e1-e2.
This system, which works so well at controlling p1 and p2, is placed into conflict
(and loses control of both p1 and p2) if p1 and p2 become non orthogonal (highly
correlated with each other). The easiest way to do this is to change p1 so that it
equals x1-x2, just like p2, or change p2 so that it equals x1 + x2, just like p1.
This change in the perceptual variables controlled instantly places the control
systems into conflict, with no changes in the output functions. Indeed, there is
no way to change the output functions so as to "fix" the conflict. That is why I
say that the conflict is defined in terms of _perception_ not _output_.
I hope this helps. But I do suggest that you test these things out for yourself.
You could use my spreadsheet to test it. You can easily create a conflict in the
spreadsheet by making two systems at the same level control the same variable (or
variables that are highly correlated, ie., have one control 3x+4y and another
control 2x+7y).
Best regards
Rick
---
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org
[From Bruce Gregory (2003.0630.1514)]
[From Bill Powers (2003.05.30.0637 MDT)]
The way I have it worked out, emotion is what we experience after an error signal has led to preparing the somatic system for action. It might enter shortly after you have looked up to see the car ahead only two car lengths away and approaching at an accelerating rate, and are in the process of jamming the brake through the floor. You may not actually feel any emotion until well after you've saved yourself from a wreck.
Why do you jam the brake through the floor? In other words, why is the gain so high on this particular control loop at this particular time? Is the gain always this high? What does this say about the control of distance between you and other cars? Why don't you find yourself slamming on the accelerator and brake alternately if the gain is always this high? Or is the gain a function of error? What does that say about how we brake in an emergency? I don't know about you, but I brake very hard very early in the process.
As I read the above 'emotion' is an epiphenomenon - a ghost in the control machine. Is that correct?
[From Bruce Gregory (2003.0630.1526)]
Rick Marken (2003.06.30.1140)
Conflict occurs when two systems at the same level in the hierarchy
(two systems
controlling the same types of perceptual variables) are controlling
non-orthogonal
(mathematically equivalent) perceptual variables relative to different
references. Conflict is a perceptual, not an output, phenomenon.
Conflict exists
when two control systems act to keep what is virtually the same
perceptual
variable in two different states. In order to do this, the two systems
are
_usually_ (depending on disturbances) trying to get the same output
into two
different states.
I follow you up to this point.
But this is not what defines the conflict.
What _does_ define the conflict, then?
In a hierarchical
system, higher level systems are almost always trying to get the same
lower level
outputs into different states (in the sense that the reference for the
lower level
systems is the sum of the outputs of several higher level systems).
I'm truly lost. I guess I don't understand conflict in the sense the
term is used in HPCT.
[From Bill Powers (2003.05.30.0637 MDT)]
David Goldstein (2003.06.30 EDT)
Bill,
After surviving my airplane crash in 1974,
I remember going into situations and scanning the environment for any
indications of dangerousness.
If I were driving in a car as a passenger,
I remember that the experience of small space, and the
scene moving by were enough to trigger an emotion of fear. I remember pushing
my foot against the floor, even when I was a passenger. I remember holding onto
the dashboard as if to brace myself.
All of this was based on one-trial
learning from the airplane crash. Pavlovian kind of conditioning seemed to have
taken place. This way of experiencing cars was not present before. It took
place after the crash and was not based on any current reality.
I have a patient who was rear-ended while
driving his car. He has a metal contraption in his back to provide support
while his back is healing. It has been 15 months post accident. He continues to
experience pressure and pain from his back. Recently, we discovered while
having him hooked up to an EMG unit, that he tightened his muscles when sitting
down and standing up for longer than he had to in order to achieve the result. This was
causing him unnecessary pain and discomfort. He became aware of the thought
that he was always bracing for the next hit. With the new awareness, he makes
an effort to relax his back muscles after the movement is achieved. He says
that this helps to reduce the pain to some degree.
At what level of the control system hierarchy
does emotion first emerge?. I think that the sensation
level may be the lowest level. The experience we are having can be visual and
emotional. It could be auditory and emotional. It could be a smell and
emotional. This is way before
I may engage in some kind of action. I seem to remember that by observing
the dilation of a person’s pupils, we can tell if a person likes what
they are seeing or doesn’t like it.
Bill, what is all this stuff about
hand-waving? If you are hand-waving, then all I can say is that hand-waving is
OK.
David
-----Original Message-----
From: Control Systems Group
Network (CSGnet) [mailto:CSGNET@listserv.uiuc.edu] On Behalf Of Bill Powers
Sent: Monday, June 30, 2003 10:01 AM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Ruminations on
Importance
[From Bill Powers (2003.05.30.0637 MDT)]
David Goldstein (2003.06.30 EDT)
I know that I am a Johnny-come-lately to the
discussion. However, I think that Bruce is bringing up some interesting issues.
What is the role of emotion when it comes to the question of what we pay
attention to?
The way I have it worked out, emotion is what we experience after an error
signal has led to preparing the somatic system for action. It might enter
shortly after you have looked up to see the car ahead only two car lengths away
and approaching at an accelerating rate, and are in the process of jamming the
brake through the floor. You may not actually feel any emotion until well after
you’ve saved yourself from a wreck.
Do you agree on this sequence, or do you, too, claim that the amygdala can
sense the visual image of the car ahead, interpret it correctly, and respond to
it by sending signals to the muscles that extend the leg against the brake? Do
you claim that the emotion arises directly from the image of the car ahead and
is the cause of the amygdala’s response? If so, join the ranks of those who
accuse me of arm-waving, while madly flailing their own around.
I would agree that attention seems drawn to, or seeks out, systems in which
unusually large error exists. Also, reeorganization seems to be focused where
attention is. Causality is not clear in all this.
Bill, you mention that you notice that your muscles
become tense when the traffic is heavy.
I proposed that they may become tense as a way of increasing loop gain, since
the stiffness of active muscle is greater than that of flaccid muscle (the
force-stretch curve is in fact exponential). I definitely don’t think that
heavy traffic causes my muscles to become tense. Higher-level control systems
do.
Isn t this because you are afraid that the chances of
an accident are higher and that you don t want to be in an accident with all
that implies?
I doubt that cognitive reasoning comes into it except on a very slow time scale
prior to any emergencies and after the fact as we attempt to explain our
behavior when the emergency is over. I would guess that when errors start to
get larger as traffic density increases , higher systems experience somewhat
greater errors and raise the gain (perhaps by using the amygdala if it is
connected in a way that makes that feasible) to limit the errors (proximity
errors, in traffic). There are other ways of increasing loop gain which are
probably also used (see above). The gain increase could simply be a consequence
of an increased error signal entering a nonlinear output function.
Might not this be a case where the Observer enters the
picture?
I think the Observer is always in the picture except perhaps when you’re
unconscious. It’s not clear to me that the Observer and attention are necessarily
the same thing, especially when attention is measured by such things as
direction of gaze. Sitting in a movie with your girlfriend, your eyes are on
the screen but your attention may well be on the girl.
Yes, I want to get to the meeting on time. Yes, I want
to advance my career.
But my life is more important than all of that.
No conflict, I put my foot on the brake.
I will be late, but I live to fight another day.
By that time, your front bumper is embedded in the trunk of the car ahead. I
think all these considerations are better conceived of as simultaneous control
processes, not a series of steps in after-the-fact verbal reasoning. You are
proceeding toward the meeting, or work. You are furthering your career. You are
making a living. You are avoiding dangers. These are all perceptual
consequences of whatever varying mix of actions you’re carrying out at the
moment. Error signals of some size exist in all these systems at the same time,
and contribute to the settings of reference signals at the next lower order,
whatever that may be for the system in question.
One of the lower systems, it stands to reason, is concerned with collision
avoidance. The reference level for minimum acceptable distance may be set by a
higher system, and may vary with traffic density (but because of the resulting
proximity errors, not because of the traffic density “stimulus”).
Given the reference setting, the actions involved in collision avoidance depend
only on the error in the system controlling perceived distance. If the distance
is too small, the collision avoidance system will start applying the brakes,
with a speed depending on the amount of error. This will not cause any
appreciable change in the error signal at the level of “get to the meeting
on time,” because the higher variable (say, estimated time of arrival)
will not change much, percentagewise, on a time scale of a few seconds. The
speed at which you’re driving is already a compromise among several different
control systems. Neither a few seconds of braking or a few seconds of
accelerating will change it much on the time scale, and the perceptual scale,
of the higher system. So an episode of sudden braking to avoid a collision will
be followed by an acceleration back to an acceptable speed, and perhaps by a
higher-order decision to terminate the cell-phone conversation to keep your
gaze on the road and your free hand on the steering wheel instead of gesturing
to emphasize a point. This episode will have only a minor effect on the
perceived time of arrival, and will be finished much too quickly for the ETA
system to affect it while it’s in progress.
That’s my story, and I maintain that it hangs together better, and involves a
whole lot LESS arm-waving, than the other story, though I will admit to extending
a hand once in a while – just to steady myself, you understand.
You’re a clinical psychologist: see if you can guess what else I might be
restraining myself from saying.
Best,
Bill P.
[From Bruce Gregory (2003.0701.0505)]
Rick Marken (2003.06.30.1450)
I hope this helps. But I do suggest that you test these things out for yourself.
You could use my spreadsheet to test it. You can easily create a conflict in the
spreadsheet by making two systems at the same level control the same variable (or
variables that are highly correlated, ie., have one control 3x+4y and another
control 2x+7y).
Thanks. It definitely does help and I will test it myself with your
spreadsheet.
--
Bruce Gregory lives with the poet and painter Gray Jacobik in the future
Canadian Province of New England.
www.joincanadanow.org
[From Bill Powers (2003.07.01.2112 MDT)]
Bruce Gregory (2003.0630.1514)--
Why do you jam the brake through the floor? In other words, why is the
gain so high on this particular control loop at this particular time?
Increasing gain does not normally result in increasing output; it results
in reducing error (making control more precise). When you maintain
reasonably precise spacing between your car and the car ahead, control is
achieved by small variations in foot pressure driven by small variations in
the spacing relative to your reference level for spacing. The higher your
gain, the less error it will take to produce enough change in output to
counter any variations. This means that errors decrease as gain increases,
while changes in action simply oppose disturbances a little more exactly.
If the gain was 10, then the output would be varying 10/11 as much as the
disturbance; if the gain increases to 100, the output will vary 100/101 as
much as the disturbance -- an increase in output of only about 10%. But the
error will decrease by a factor of 10, from 10% of the reference value to
1% of the reference value.
Suppose that when you follow a car with a 30-foot separation, you manage to
keep the distance variations down to plus or minus 3 feet -- 10% of the
spacing. You do this mainly by varying acceleration (accelerator pedal
force). If you start overtaking the car ahead so fast that taking your foot
off the accelerator will not correct the distance error, you start to apply
the brake. The amount of pressure you apply to the brake depends on (1) how
much too close you are, and (2) the gain of your control system.
In several other control tasks like this, human performance is
well-represented by a control system that controls a combination of the
primary variable and its first derivative. Adding the first derivative
means in this case that the perception of the rate of change of distance to
the car ahead -- the relative velocity -- is added to the perception of
distance, so the system reponds to changes with a higher gain than to
steady-state values. This type of perceptual reponse is typical of many
lower-order perceptions such as distance perception.
So how hard you apply the brake also depends on a third variable: the rate
of change of perceived distance. While you're looking away from the road,
you have to imagine that the distance to the car ahead is remaining the
same -- otherwise you might speed up as soon as you looked away, since the
car ahead would seem to have disappeared from view. And when you look up
again to see a diminished distance, substituting the real perception for
the imagined one, there would be a very sudden decrease in perceived
(distance + change of distance) because the derivative of a step-change is
very large. You would therefore step on the brake very hard, but only long
enough to reduce the error to a small value. The loop gain is the same as
it always was, but the error signal is much larger than the control system
normally permits.
Of course that's merely an attempt to propose a way in which a PCT model
could explain the phenomenon you describe. To test this proposal would
require experiments with drivers and instrumented cars.
Is the gain always this high? What does this say about the control of
distance between you and other cars?
It says that you normally keep the perceived separation fairly close to the
separation you want.
Why don't you find yourself slamming on the accelerator and brake
alternately if the gain is always this high?
Because your control system is stable and doesn't "hunt".
Or is the gain a function of error? What does that say about how we
brake in an emergency? I don't know about you, but I brake very hard very
early in the process.
Yes, because the error gets much larger than is usually the case. If you're
interested, I'll send you a model of a driver with a distance + rate of
change control system braking a car to a stop at a preselected position
(related to the situation you describe). Bruce Abbott worked out a
particularly nice Delphi version based on a cruder one of mine. This model
is part of an ongoing interaction with John Flach and some other
engineering psychologists at Wright State College in Dayton, OH, so is not
to be cited yet. In this model, the driver exerts a braking force, with the
twist that there is an auxiliary control system that reduces the sensed
deceleration if it exceeds a specific value (a one-way control system). The
result is that there is hard but limited braking at first, as if stopping
on a wet road or ice, followed by a "creep" to the final position (the gain
of the deceleration-limiting system can be varied or even dropped to zero,
to see what happens when there is no limit). If the initial velocity is too
high or the stopping point chosen is too close, the braking goes to maximum
but the car overshoots the target position. All the parameters can be
varied by dragging sliders with the mouse, so you can simulate various
conditions. You could easily set up a case where the driver is suddenly
presented with a stopping position that is very close, which would be at
least similar to the case you describe.
As I read the above 'emotion' is an epiphenomenon - a ghost in the control
machine. Is that correct?
I do not believe in ghosts.
An epiphenomenon is something caused by brain activity but which has no
ability to affect that activity. The term is sometimes applied (mistakenly,
I think) to consciousness, or "mental" activities. Emotion is not an
epiphenomenon in my model; it is a combination of somatic sensations with
perceived states of activity in higher-level control systems, both being
perfectly real physical phenomena. Large or protracted error signals in the
hierarchy result not only in driving energetic actions by lower systems,
but in altering the biochemical state of the body, probably through paths
involving the hypothalamus and neurohypophysis (neural part of the
pituitary). The hypothalamus may be activated by the amygdala as well as by
other structures. Thus emotion is a hybrid phenomenon that arises from
preparation to act, and thus ultimately from error signals in the hierarchy.
Behind emotions are your reference signals and any errors that exist
between perception and reference. What you intend or want, and your success
in achieving your intentions and wants, is what creates emotion. That's the
model. I think it fits the facts.
Best,
Bill P.
[From Bruce Gregory (2003.0702.0507)]
Bill Powers (2003.07.01.2112 MDT)
Behind emotions are your reference signals and any errors that exist
between perception and reference. What you intend or want, and your success
in achieving your intentions and wants, is what creates emotion. That's the
model. I think it fits the facts.
And if you "turned off" the emotional system, the hierarchy would
control exactly as it does with the emotional system "turned on?" That
is what I imprecisely called being an epiphenomenon. I think that for
some people consciousness "functions" in exactly this way, whatever you
call it.
--
Bruce Gregory lives with the poet and painter Gray Jacobik in the future
Canadian Province of New England.
www.joincanadanow.org
[From Bill Powers (2003.07.02.0726 MDt)]
David Goldstein (2003.06.30 EDT)--
After surviving my airplane crash in 1974, I remember going into
situations and scanning the environment for any indications of dangerousness.
From that I would guess that you did not want to experience another
episode of that kind and were controlling for absence of any situation you
would categorize as dangerous.
If I were driving in a car as a passenger, I remember that the experience
of small space, and the scene moving by were enough to trigger an emotion
of fear. I remember pushing my foot against the floor, even when I was a
passenger. I remember holding onto the dashboard as if to brace myself.
My theory of emotion would say that this "triggering" of fear was not
direct, but occurred because you wanted to avoid being confined in a small
space, and to keep the car from going too fast, and to keep yourself from
being thrown against the barrier in front of you (in case of a crash).
These were things you wanted very much to avoid, for good reasons, yet for
some reason you could not (or would not let yourself) avoid them. So you
were prepared to avoid them, both in the sense of motor preparation and in
the sense of somatic (biochemical) preparation. What you experienced as
fear was a mixture of the cognitive desire to avoid pain or injury, and the
physiological preparation to support the pending action (which never, of
course, really occurred). The VERY SAME feelings, coupled to other goals
such as a desire to attack or push something away, might be called "anger"
or "outrage." You perceive the whole pattern as fear because it is coupled
with a desire to escape or avoid something.
All of this was based on one-trial learning from the airplane crash.
Pavlovian kind of conditioning seemed to have taken place. This way of
experiencing cars was not present before. It took place after the crash
and was not based on any current reality.
I don't think that "Pavlovian conditioning" explains anything. I'd rather
put it this way: you spent a considerable period of time wanting to avoid
crashing to the ground in an airplane, being smashed agonizingly against
the seat in front of you, and ending your life. All during this time,
adrenaline was pumping into your bloodstream, your heart was pounding, you
had a horrible feeling in your solar plexus, your breathing was shallow,
your skin clammy, and your muscles were tensed -- and there wasn't a damned
thing you could to do control the outcome. So something in you decided that
this set of perceptions, or any perceptions that seemed in the same
category, were to be avoided as energetically as necessary. Maybe this was
reorganization, or maybe it's just a logical process we do automatically.
The result was that any time this control system perceived something that
fell into the "danger" category, it would attempt immediately to avoid
letting those perceptions get any stronger, and to eliminate them. This
would involve not only acting physically, but cranking up the body to
support energetic or even violent action. If nothing prevented the action,
that is what you would do. If going too fast you would apply the brakes as
hard as necessary; if trapped in a confined space you would shove with all
your might and break out of it; if you were about to crash you would shove
against the barrier in front of you, preferring a broken arm to a smashed face.
However, you also recognize, rationally, that in many situations where such
perceptions occur, there is really not much actual danger, and anyway that
you aren't the driver and have no control of the situation. So when this
control system prepares to escape, brake, or brace, your rational control
system arranges for another control system to cancel out these actions by
setting up a direct conflict. You don't want to look foolish, or like a
sissy, or as though you're panicking, so you don't act. But your body is
prepared anyway, and you feel the adrenaline, the shallow breaking, the
sensation in the solar plexus, the clammy skin, and the racing heart --
just as you did in the airplane, because it's the same somatic preparation
for action. And what do you call this combination of desire to escape with
those bodily feelings? Fear, of course.
At what level of the control system hierarchy does emotion first emerge?.
I think that the sensation level may be the lowest level. The experience
we are having can be visual and emotional. It could be auditory and
emotional. It could be a smell and emotional. This is way before I may
engage in some kind of action.
My theory of emotion says that these feelings ARE preparations for action,
and that if you could act, you would be doing so even as the feelings are
arising. What we call emotions are the internal sensations from our bodies
resulting from whatever goal, at whatever level of the hierarchy, is to be
achieved. The only reason you're not acting is that there is nothing you
can do to achieve control, or because for your own reasons you will not
allow yourself to take the necessary action. This means that the primary
experience is one of being physically prepared to act to correct an error
or achieve a goal, while in fact you're not acting at all.
John Flach said something profound in an offline post. He said that
psychology has never recognized perception as a real phenomenon as we do in
PCT. Instead, it has tried to explain behavior by referring to cues, hints,
suggestions, intimations, subliminal stimuli, and other fuzzy vague sorts
of things. It's that sort of concept that leads to proposing that a smell
stimulates the olfactory bulb which stimulates the amydala which stimulates
the hypothalamus which stimulates the pitutary, which makes the heart pound
and the adrenaline rush and so forth -- all before we know consciously that
anything at all is happening. I think that's a pretty feeble story, and
that it makes much more sense to start with actual perceptions and goals
for them, which are the primary features of our conscious existence and not
subtle hints, and figure out how the process of correcting error can
involve both the motor systems and the organ systems. You end up including
the same basic signal pathways, but you don't require this or that little
part of the brain to have all the perceptions, interpretive abilities, and
control processes of the entire brain. The traditional concept of emotion,
where it manages to propose anything substantive at all, is full of logical
holes.
Bill, what is all this stuff about hand-waving? If you are hand-waving,
then all I can say is that hand-waving is OK.
Handwaving happens at point in an argument where you hear someone say, for
example, "It's not this or that, it's just, sort of, you know [wave arms
here] THE WHOLE THING." We all do it. Lately it seems that neuroscientists
are doing it a lot, especially when explaining what it means when a little
region of a brain lights up in a brain scan. Nobody can really figure out
what the parts of the brain do from information like that, but people like
to give explanations, so they just wave their arms and say, "It's caused
by, sort of, you know, THE AMYGDALA." And they show a picture of a fuzzy
blob superimposed on a slice of brain, to prove it.
Best,
Bill P.