Love and Hate

Bruce_Gregory4 · December 7, 2003, 2:06pm

[From Bruce Gregory (2003.12.07.0905)]

While I like Rick's morality tale that I have a love/hate relationship
with PCT, I'm not sure there is enough there to either love or hate.
PCT is a simple model of purposeful behavior. To the extent you can
make use of it, it is a superior tool. All the messy neurology and
networking is hidden in the perceptual function, and that's a good
thing, because we have no useful way to model this complexity--yet.

HPCT is somewhat more problematic. I've always been bothered by the
question as to how the hierarchy decides what goals to pursue and does
so quickly enough to be adaptive. This process involves, as far as I
can see, two mechanisms. Reference levels established from above and
same-level systems with higher gain dominating systems with lower
gain. Beyond that there seems little more to say at this time.

Economics can certainly use good models, but until HPCT provides a more
detailed picture of perceptual functions and the role they play in
cooperative behavior, I suspect little progress with be made from the
control theory perspective.

I'm not sure why this description strikes anyone as involving a
conflict between love and hate, but I can live with that.

Remember Pearl Harbor!

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bill_Williams1 · December 7, 2003, 6:34pm

Bruce,

I think your assessment of the implication of control theory for economics is valuable because of the extreme simplification to which you reduce matters. Understand that I'm not saying, not yet anyway, that in reducing issues thisway you are making a mistake, or mistakes.

However, you say,

Economics can certainly use good models, but until HPCT provides a more
detailed picture of perceptual functions and the role they play in
cooperative behavior, I suspect little progress with be made from the
control theory perspective.

I find your "suspicion" surprizing. As best I can tell you have good understanding of the principles of control theory and yet you see the relationship between the present state of application of control theory to human behavior as such that it generates the conclusion that "little progress" will be made until there is a marked advance in the sophistication of the control theory analysis of human perceptual functions.

In contrast from my perspective the possiblities for a control theory based economics look somewhat different. Over the years I've managed, initially in colaboration with Powers, and then independently as I gained some elementary program skills, to create control theory models of the 1) Giffen paradox, 2) a Giffen curve, and a 3) Giffen surface, 4) dynamic Giffen models, a 5) global model of a consumer from a control theory standpoint 6) a model of behavior known as the Veblen/Dussenberry effect, 7) the backward bending labor curve,
8) Veblen's model of Conspicious Consumption, 9) the blue-jeans effect ( a reverse of the conspicious consumption effect ), 10) an update in control theory terms of Burridan's Ass paradox, 11) an information flow demo that in a lattice shows how a delay loop generates profound instablities, this model has application both in economics to questions where there is a differential access to information, and in political science in regard to a substansive as contrasted to a formalist theory of democracy. 12) an application of Powers' multiple control work to the problem known as the theory of institutional adjustment, 13) the recent two commodity model, and a 14) single good control theory model of demand upon which it is based, and a control theory interpretation of what are known as 15) hurdle rates, a concept in the theory of the firm under the caption of 16) X-efficiency, a re-interpretation of Keynes' model of the economy in which the fundamental variables are specified in control theory terms, and 17 a re-interpreation of the pragmatism in philosophy, and a similar re-reading of the work of 18) Veblen, and 19) Commons in economics in terms of control theory, plus an intepretation of 20) Foster's conception of institutional adjustment in a control theory perspective. And, last 21) a fulfillment of Bouldings prediction that an application of cybernetics to economics would revolutionize the conception of the economic process. 22? ) there's Powers economic testbed that is already partly in operation. And, 23?? ) it just recently occurred to me that the various descriptions of the theory of the firm and competition could be rewritten in control theory terms. I've tended to think such rewriting exercises would be trival and not worth the effort. However, after completing (more or less) the two commodity model I've changed my mind in this regard. 24??) it seems to me that it would be possible to generate a theory of advertising and salesmanship from a control theory standpoint. If you inspect a contemporary economics introductory text you won't find an explaination of either advertising or salesmanship-- this is I think most peculiar. There is also a control theory explalination for are known as 25) Veblenian price effects.

Understand I'm not claiming that above examples necesarily amount to a comprehesive alternative model of the economy, but at least in my own view the examples are sufficiently numerous that one might be persuaded that over time and with more hands involved it would be possible to create a viable model of the economy which would be based upon a control theory foundation. And, as best I can see a realization of such an alternative economic conception would not require further developments in control theory in order to be carried out.

I wish to thank you for providing me with what amounts to what I regard as an ideal audience-- that is an audience that is sufficiently informed concerning a basic context, yet skeptical concerning the argument involved. Simply listing the work already done, as in the above section, is something I've never actually done-- not in anything approaching the way I've done here in the listing. After all why should I write down this stuff when its all in my head? But, the neccesity of writting it down in the process of attempting to communicate to a skeptical audience has made for a marked change in the way I perceive it myself.

Bill Williams

···

-----Original Message-----
From: Control Systems Group Network (CSGnet) on behalf of Bruce Gregory
Sent: Sun 12/7/2003 8:06 AM
To: CSGNET@listserv.uiuc.edu
Subject: Love and Hate

[From Bruce Gregory (2003.12.07.0905)]

While I like Rick's morality tale that I have a love/hate relationship
with PCT, I'm not sure there is enough there to either love or hate.
PCT is a simple model of purposeful behavior. To the extent you can
make use of it, it is a superior tool. All the messy neurology and
networking is hidden in the perceptual function, and that's a good
thing, because we have no useful way to model this complexity--yet.

HPCT is somewhat more problematic. I've always been bothered by the
question as to how the hierarchy decides what goals to pursue and does
so quickly enough to be adaptive. This process involves, as far as I
can see, two mechanisms. Reference levels established from above and
same-level systems with higher gain dominating systems with lower
gain. Beyond that there seems little more to say at this time.

Economics can certainly use good models, but until HPCT provides a more
detailed picture of perceptual functions and the role they play in
cooperative behavior, I suspect little progress with be made from the
control theory perspective.

I'm not sure why this description strikes anyone as involving a
conflict between love and hate, but I can live with that.

Remember Pearl Harbor!

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

bnhpct · December 7, 2003, 6:48pm

[From Bruce Nevin (2003.12.07 13:48 EST)]

Bruce Gregory (2003.12.07.0905)–

All the messy neurology and

networking is hidden in the perceptual function,

Actually, there is messy complexity under the rug at all three interior
points of the loop: perceptual input to the comparator, reference input
to the comparator, and the output function that combines error signals
and sends output signals to effectors. And outside in the environment
there are various sorts of feedback as soon as there are other control
systems participating. (Possibly loops through non-living interactions
too – at least Bateson thought so.) All four corners of the rug,
then.

and that’s a good

thing, because we have no useful way to model this
complexity–yet.

HPCT is somewhat more problematic. I’ve always
been bothered by the

question as to how the hierarchy decides what goals to pursue and
does

so quickly enough to be adaptive.

This has been discussed but is not presently shown by any simulation that
I know of (I could be wrong). In such a simulation, the same effector is
at the employ of more than one higher-level purpose. I can think of three
ways of doing this without conflict that have been discussed:

Under sequence or program control, where conflict does not arise.
With sufficient difference of gain to decide the conflict.
With rapid conflict resolution by established problem-solving
routines.

This process involves, as far as I

can see, two mechanisms. Reference levels established from above

That is, (1) and (3), possibly others.

and

same-level systems with higher gain dominating systems with lower

gain. Beyond that there seems little more to say at this
time.

There’s a limit to the value of talk about it. The Crowd simulation
demonstrates (2). Is there any simulation that demonstrates (1) or
(3)?

Economics can certainly use good models, but
until HPCT provides a more

detailed picture of perceptual functions and the role they play in

cooperative behavior, I suspect little progress with be made from
the

control theory perspective.

Here’s my idea of what stops us. It’s possible to abstract the lower
levels and simulate agents who inhabit a universe of relationships,
sequences, etc. among categories. However, such a simulation is
unconvincing because unavoidably the categories are determined a priori
by theoretical preconceptions about the relationships, sequences, etc. of
economics. Relevant categories are determined a priori, yes, but
they are not created by the theoretical pronouncements and debates of
pundits, which after all are supposed to be descriptive. Descriptive of
what? Are these categories all cultural values that are established and
recreated in the social situations in which agents control them, as the
conventions of language are? How else could they arise? (That is not a
rhetorical question.)

    /Bruce

Nevin

···

At 09:06 AM 12/7/2003 -0500, Bruce Gregory wrote:

Richard_Marken2 · December 7, 2003, 7:34pm

[From Rick Marken (2003.12.07.1130)]

Bruce Gregory (2003.12.07.0905)

HPCT is somewhat more problematic. I've always been bothered by the
question as to how the hierarchy decides what goals to pursue and does
so quickly enough to be adaptive.

The spreadsheet hierarchy shows how this is done. I suggest
downloading it and taking a look at how the references at each level
change over time as disturbances in the environment continuously vary.
It's a very pretty process, really.

This process involves, as far as I
can see, two mechanisms. Reference levels established from above and
same-level systems with higher gain dominating systems with lower
gain. Beyond that there seems little more to say at this time.

The relative gain of same-level systems is not really part of the
process of hierarchical control. I would say that the most important
aspects of the hierarchical control process are the following: 1)
Different _types_ of perceptual variables must be controlled by each
level of the hierarchy and 2) All control systems at the same level
must control independent degrees of freedom of perceptual variables
from the next lower level. In the spreadsheet, the different types of
perceptual variables are defined by the types of the perceptual
functions that transform lower level inputs into perceptual variables.
At the lowest level, the type of function (perception) is linear: eg. p
= ks. At the next level the type of function (perception) is a vector:
eg. p2 = k1p1+k2p2...k6p6. At the highest level the type of function
(perception) is a logical relationship: eg. if p21>p22, p3=1, else
p3=0. The spreadsheet is currently set up so that all perceptual
functions at all levels control independent degrees of freedom in the
array of perceptions (or sensory inputs) from the next lower level. If
you change the perceptual functions so that two systems at the same
level control the same perception (same degree of freedom) control by
the two systems and any higher level systems that use those systems as
the means of controlling their own perceptions is lost.

Economics can certainly use good models, but until HPCT provides a more
detailed picture of perceptual functions and the role they play in
cooperative behavior, I suspect little progress with be made from the
control theory perspective.

The spreadsheet model shows how N systems at one level of the hierarchy
can cooperatively set the references for N lower level systems in order
to continuously achieve their goals in the face of continuously
changing disturbance. HPCT shows that this kind of cooperation can be
achieved only when the N higher level systems control perceptions that
are orthogonal. So HPCT already provides a detailed picture of the
role of perceptual functions in cooperative behavior. The picture is
provided by mathematics rather than neurology. The mathematics
(implemented as a working model in the spreadsheet) show that
cooperation and conflict are two ends of a spectrum defined by the
degrees of freedom available for control.

Best regards

Rick

···

---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Bruce_Gregory4 · December 7, 2003, 7:57pm

[From Bruce Gregory (2003.12.07.1456)]

[From Rick Marken (2003.12.07.1130)]

Bruce Gregory (2003.12.07.0905)

HPCT is somewhat more problematic. I've always been bothered by the
question as to how the hierarchy decides what goals to pursue and does
so quickly enough to be adaptive.

The spreadsheet hierarchy shows how this is done. I suggest
downloading it and taking a look at how the references at each level
change over time as disturbances in the environment continuously vary.
It's a very pretty process, really.

I'm sure it is. But is it a realistic model of my traffic example?

This process involves, as far as I
can see, two mechanisms. Reference levels established from above and
same-level systems with higher gain dominating systems with lower
gain. Beyond that there seems little more to say at this time.

The relative gain of same-level systems is not really part of the
process of hierarchical control. I would say that the most important
aspects of the hierarchical control process are the following: 1)
Different _types_ of perceptual variables must be controlled by each
level of the hierarchy and 2) All control systems at the same level
must control independent degrees of freedom of perceptual variables
from the next lower level. In the spreadsheet, the different types of
perceptual variables are defined by the types of the perceptual
functions that transform lower level inputs into perceptual variables.
At the lowest level, the type of function (perception) is linear: eg. p
= ks. At the next level the type of function (perception) is a vector:
eg. p2 = k1p1+k2p2...k6p6. At the highest level the type of function
(perception) is a logical relationship: eg. if p21>p22, p3=1, else
p3=0. The spreadsheet is currently set up so that all perceptual
functions at all levels control independent degrees of freedom in the
array of perceptions (or sensory inputs) from the next lower level. If
you change the perceptual functions so that two systems at the same
level control the same perception (same degree of freedom) control by
the two systems and any higher level systems that use those systems as
the means of controlling their own perceptions is lost.

So it _doesn't_ model my traffic example. That's not a criticism, just
an example of situations that are not yet modeled by HPCT.

Economics can certainly use good models, but until HPCT provides a
more
detailed picture of perceptual functions and the role they play in
cooperative behavior, I suspect little progress with be made from the
control theory perspective.

The spreadsheet model shows how N systems at one level of the hierarchy
can cooperatively set the references for N lower level systems in order
to continuously achieve their goals in the face of continuously
changing disturbance. HPCT shows that this kind of cooperation can be
achieved only when the N higher level systems control perceptions that
are orthogonal. So HPCT already provides a detailed picture of the
role of perceptual functions in cooperative behavior.

Actually, I was referring to cooperation involving different
hierarchies. For example, when one person perceives that others are
buying a stock, thus making it more attractive.

The picture is
provided by mathematics rather than neurology. The mathematics
(implemented as a working model in the spreadsheet) show that
cooperation and conflict are two ends of a spectrum defined by the
degrees of freedom available for control.

I agree. Provided the perceptual variables are established in the way
you specify. I suspect this applies to a relatively restricted set of
problems in the real world, however. It clearly does not apply to my
traffic example.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bruce_Gregory4 · December 7, 2003, 8:09pm

[From Bruce Gregory (2003.12.07.1505)]

···

On Dec 7, 2003, at 1:34 PM, Williams, William D. wrote:

I wish to thank you for providing me with what amounts to what I
regard as an ideal audience-- that is an audience that is sufficiently
informed concerning a basic context, yet skeptical concerning the
argument involved. Simply listing the work already done, as in the
above section, is something I've never actually done-- not in anything
approaching the way I've done here in the listing. After all why
should I write down this stuff when its all in my head? But, the
neccesity of writting it down in the process of attempting to
communicate to a skeptical audience has made for a marked change in
the way I perceive it myself.

Thank you. I stand much better informed. I'll have to re-examine my
prejudices about PCT and economics.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Richard_Marken2 · December 7, 2003, 8:25pm

[From Rick Marken (2003.1207.1225)]

Bruce Gregory (2003.12.07.1456)--

Rick Marken (2003.12.07.1130)--

Bruce Gregory (2003.12.07.0905)

HPCT is somewhat more problematic. I've always been bothered by the
question as to how the hierarchy decides what goals to pursue and
does
so quickly enough to be adaptive.

The spreadsheet hierarchy shows how this is done. I suggest
downloading it and taking a look at how the references at each level
change over time as disturbances in the environment continuously vary.
It's a very pretty process, really.

I'm sure it is. But is it a realistic model of my traffic example?

The spreadsheet answers your question about how a control hierarchy
"decides what goals to pursue and does so quickly enough to be
adaptive". The spreadsheet was designed as a teaching tool, to
illustrates the _principle_ of hierarchical control. It was not meant
to be a model of a specific behavior. (Good examples of the
application of principles of hierarchical control in models of a
specific behavior are the Crowd and Little Man programs).

So the spreadsheet hierarchy is not likely to be a realistic model of
your traffic example, whatever that is. If you have some data on
traffic behavior that seems to require a hierarchical control model I
think it would, indeed, be instructive to try to work out a realistic
model of that behavior.

Best regards

Rick

···

---

Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Bruce_Gregory4 · December 7, 2003, 8:42pm

[From Bruce Gregory (2003.12.07.1540)]

Rick Marken (2003.1207.1225)

So the spreadsheet hierarchy is not likely to be a realistic model of
your traffic example, whatever that is. If you have some data on
traffic behavior that seems to require a hierarchical control model I
think it would, indeed, be instructive to try to work out a realistic
model of that behavior.

Fair enough. My traffic example is excruciatingly simple. I am driving
to work. My foot is on the accelerator. A car cuts in front of me. I
slam on the brakes. My reaction is too rapid for conflict to be
resolved by higher level systems adjusting goals. This seems to be a
case, like Crowd, where one system has sufficient gain to overpower
other systems on the same level ("position of my foot"). If I
understand you rightly, this situation is not part of HPCT model, at
least as far your demonstration is concerned.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Richard_Marken2 · December 7, 2003, 9:16pm

[From Rick Marken (2003.12.07.1315)]

Bruce Gregory (2003.12.07.1540)]

Rick Marken (2003.1207.1225)

So the spreadsheet hierarchy is not likely to be a realistic model of
your traffic example, whatever that is.

Fair enough. My traffic example is excruciatingly simple. I am driving
to work. My foot is on the accelerator. A car cuts in front of me. I
slam on the brakes.

This seems pretty straight forward.

My reaction is too rapid for conflict to be
resolved by higher level systems adjusting goals.

So?

This seems to be a
case, like Crowd, where one system has sufficient gain to overpower
other systems on the same level ("position of my foot").

Very likely.

If I understand you rightly, this situation is not part of HPCT model

You lost me. You just said that it sounds to you like the model of this
situation would be equivalent to the model of an agent in Crowd. Each
agent in Crowd is modeled as a hierarchy of control systems. So you
apparently realize that the situation can be handled by HPCT. What's
the problem?

, at least as far your demonstration is concerned.

Yes. Of course. I said that my spreadsheet is a tutorial demonstration
of principle that doesn't pertain to any specific situation. It
illustrates principles of hierarchical control that can be used to
model specific behavioral situations. I think it would be pretty easy
to use those principles in a two level spreadsheet model of a driver
who drives to a destination (work) and doesn't hit the cars in front
(which are braking and accelerating unpredictably) while doing so.
Actually, Bill has already modeled something very close to this
situation in the Crowd program.

Best regards

Rick

···

---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Bruce_Gregory4 · December 7, 2003, 9:47pm

[From Bruce Gregory (2003.12.07.1645)]

[From Rick Marken (2003.12.07.1315)]

You lost me. You just said that it sounds to you like the model of this
situation would be equivalent to the model of an agent in Crowd. Each
agent in Crowd is modeled as a hierarchy of control systems. So you
apparently realize that the situation can be handled by HPCT. What's
the problem?

The problem appears to be that I believed you when you said, "The
relative gain of same-level systems is not really part of the process
of hierarchical control." Were you pulling my leg? Or do you want to
have it both ways?

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bill_Williams1 · December 7, 2003, 9:50pm

[Bill Williams 7 December 2003 3:43 PM]

Bruce,

You may remember a long ago discussion of pilot induced oscilations (PIO). For those not around then, a pilot sometimes mishandles a landing approach and the aircraft bounces harshly on the runway. As a result the pilot increases the gain in the loop so much that an oscilation results, and sometimes this oscilation can be divergent. This is an example that has been repeatedly model and it is claimed that the PIO phenomena is "well understood."

Watching student pilot entrained in PIO's I got the impression that what was going on was a conflict of a sort in which attention was shifting between controlling for not smashing the plane on the runway, and not stalling the airpline for which the reference level was either airspeed or nose attitude.

A pilot with better control of the aircraft wouldn't have experience the harsh contact with the runway. Thus they would not experience the following massive increase in gain, and a possible slowdown in perception. Both of which could destabilize the loop controlling the aircraft.

Does the PIO example relate to your guy's expereince in with the car?

Bill Williams

···

-----Original Message-----
From: Control Systems Group Network (CSGnet) on behalf of Bruce Gregory
Sent: Sun 12/7/2003 2:42 PM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Love and Hate

[From Bruce Gregory (2003.12.07.1540)]

Rick Marken (2003.1207.1225)

So the spreadsheet hierarchy is not likely to be a realistic model of
your traffic example, whatever that is. If you have some data on
traffic behavior that seems to require a hierarchical control model I
think it would, indeed, be instructive to try to work out a realistic
model of that behavior.

Fair enough. My traffic example is excruciatingly simple. I am driving
to work. My foot is on the accelerator. A car cuts in front of me. I
slam on the brakes. My reaction is too rapid for conflict to be
resolved by higher level systems adjusting goals. This seems to be a
case, like Crowd, where one system has sufficient gain to overpower
other systems on the same level ("position of my foot"). If I
understand you rightly, this situation is not part of HPCT model, at
least as far your demonstration is concerned.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Martin_Taylor2 · December 7, 2003, 10:13pm

[From Bruce Gregory (2003.12.07.1645)]

[From Rick Marken (2003.12.07.1315)]

You lost me. You just said that it sounds to you like the model of this
situation would be equivalent to the model of an agent in Crowd. Each
agent in Crowd is modeled as a hierarchy of control systems. So you
apparently realize that the situation can be handled by HPCT. What's
the problem?

The problem appears to be that I believed you when you said, "The
relative gain of same-level systems is not really part of the process
of hierarchical control." Were you pulling my leg? Or do you want to
have it both ways?

I don't understand where the situation suggests a change in _gain_.
You observe a change in output of the two systes, that occurs when
the error in the "closeness to vehicle in front" system suddenly gets
huge. That causes a huge output in a system that had no error a few
moments earlier. The output of the "getting to work on time" system
hasn't changed. Why is there a problem with the notion that the
system with a new very large output suddenly overwhelms the one whose
small (possibly zero) output hasn't changed?

Where's the change in gain implied by anything in the situation?

Martin

Bruce_Gregory4 · December 7, 2003, 10:17pm

[From Bruce Gregory (2003.12.07.1715)]

[Bill Williams 7 December 2003 3:43 PM

Bruce,

You may remember a long ago discussion of pilot induced oscilations
(PIO). For those not around then, a pilot sometimes mishandles a
landing approach and the aircraft bounces harshly on the runway. As a
result the pilot increases the gain in the loop so much that an
oscilation results, and sometimes this oscilation can be divergent.
This is an example that has been repeatedly model and it is claimed
that the PIO phenomena is "well understood."

My experience is that PIO are not examples of controlling with too much
gain, but of trying to control the wrong variables. (Having done this
myself, I feel I can speak with authority!) The must useful advice I
got was from a flight instructor who told me, "Always back; never
forward." If you control for this perception of the stick, you don't
bounce when you land.

Watching student pilot entrained in PIO's I got the impression that
what was going on was a conflict of a sort in which attention was
shifting between controlling for not smashing the plane on the runway,
and not stalling the airpline for which the reference level was either
airspeed or nose attitude.

Yes. Controlling the "wrong" perceptions.

A pilot with better control of the aircraft wouldn't have experience
the harsh contact with the runway. Thus they would not experience the
following massive increase in gain, and a possible slowdown in
perception. Both of which could destabilize the loop controlling the
aircraft.

There are certainly circumstances where excessive gain gets a pilot
into trouble. Recovery from unusual attitudes normally goes wrong
because the pilot induces stresses that the airplane cannot experience
without failure of some vital part.

Does the PIO example relate to your guy's expereince in with the car?

I don't think so. When I slam on my brakes, everything usually works
out the way I intended. If it doesn't, it probably not be because I hit
the brakes too hard. (I have anti-lock brakes.)

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bruce_Gregory4 · December 7, 2003, 10:32pm

[From Bruce Gregory (2003.12.07.1730)]

[From Bruce Gregory (2003.12.07.1645)]

[From Rick Marken (2003.12.07.1315)]

You lost me. You just said that it sounds to you like the model of
this
situation would be equivalent to the model of an agent in Crowd. Each
agent in Crowd is modeled as a hierarchy of control systems. So you
apparently realize that the situation can be handled by HPCT. What's
the problem?

The problem appears to be that I believed you when you said, "The
relative gain of same-level systems is not really part of the process
of hierarchical control." Were you pulling my leg? Or do you want to
have it both ways?

Where's the change in gain implied by anything in the situation?

There isn't any change in gain that I know of. How do you arrive at
this conclusion from the above exchange? I agree with your analysis. I
never meant to suggest otherwise.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

···

On Dec 7, 2003, at 5:13 PM, Martin Taylor wrote:

Bill_Williams1 · December 7, 2003, 10:36pm

[Bill Williams 7 December 2003 4:36 P.M.]

Bruce,

For the most part your instructor's advice will be effective. However,
in the case of the old conventional geared aircraft, and especially
the heavier two engine versions such as a Beech 18/C-45 or similiar,
it is commonly considered neccesary to move the control forward to
"stick" the aircraft on the runway. And, I too have experienced PIO's
after inept efforts to land. So recently did the pilot of a F-22
prototype-- the landing effectively destroyed the aircraft for any
further flight testing. We can assume that the pilot in command was
one of the best trained and best qualified availible for the task.

Despite this he may, as you say, been controlling for the wrong
perceptions, but I wonder. In your opinion, if the pilot always
controls for the right perceptions, can the pilot be assured that
a PIO will never occur?

Bill Williams

Richard_Marken2 · December 7, 2003, 10:42pm

[From Rick Marken (2003.12.07.1440)]

Bruce Gregory (2003.12.07.1645)--

Rick Marken (2003.12.07.1315)--

So you apparently realize that the situation can be handled by HPCT. What's
the problem?

The problem appears to be that I believed you when you said, "The
relative gain of same-level systems is not really part of the process
of hierarchical control." Were you pulling my leg? Or do you want to
have it both ways?

Neither. The sentence to which you refer was a reply to the following from you:

This process involves, as far as I can see, two mechanisms. Reference levels established from above and same-level systems with higher gain dominating systems with lower gain.

I replied by saying "The relative gain of same-level systems is not really part of the process of hierarchical control" because I don't consider relative gain to be an essential consideration when building a working HPCT model. The relative gains of systems at the same level can be quite different and still there will be successful hierarchical control (you can demonstrate this to yourself using the spreadsheet hierarchy). What is essential to the process of hierarchical control is what I mentioned earlier: that different _types_ of perceptual variable be controlled by each level of the hierarchy and that control systems at the same level control independent degrees of freedom of perceptual variables from the next lower level.

Best

Rick

···

---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

Bill_Williams1 · December 7, 2003, 11:14pm

[Bill Williams 7 December 2003 4:47 P.M]

You say,

"Are these categories all cultural values that are established and
recreated in the social situations in which agents control them,
as the conventions of language are? How else could they arise?
(That is not a rhetorical question.)"

I may misunderstand both the context and the intent of the question.
However, contrary to the assertions of the cultural relativists who
would deny the existence of any values that have anything other than
an arbitrary cultural source, there of late many people who are
seeking to locate a non-arbitrary foundation for values. I think many
of these people would say that, language included, all human behavior
can be understood in terms of a process of value creation, and that
these values are not merely conventional. The question of how such
values could arise, they might say, is identical to the question of
how could life arise.

It isn't an answer to the question but George Lichtheim 1974 _From Marx to Hegel_ says of this issue,

      "...the logical status of value judgements has become a practical
     problem not only just for a few reflective individuals, but for
     everyone." p. 214.

It seems to me that my recent two commodities control model illustrates in a mundane way how a switch to a control theory analysis can generrate a change in the value orrientation. It may among other things in fact be possible to be "too thin, and too rich."

Bill Williams

Bruce_Gregory4 · December 7, 2003, 11:19pm

[From Bruce Gregory @003.12.07.1818)]

Rick Marken (2003.12.07.1440)

I replied by saying "The relative gain of same-level systems is not
really part of the process of hierarchical control" because I don't
consider relative gain to be an essential consideration when building
a working HPCT model.

O.K. Then how does my traffic example work if one control system does
not overwhelm the other?

The relative gains of systems at the same level can be quite different
and still there will be successful hierarchical control (you can
demonstrate this to yourself using the spreadsheet hierarchy).

I think we can all agree on this.

What is essential to the process of hierarchical control is what I
mentioned earlier: that different _types_ of perceptual variable be
controlled by each level of the hierarchy and that control systems at
the same level control independent degrees of freedom of perceptual
variables from the next lower level.

Can I then assume that the systems controlling whether my foot is on
the accelerator or on the brake involve different degrees of freedom?
Both systems _seem_ to be controlling the location of my foot. What
subtlety have I failed to grasp?

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bruce_Gregory4 · December 7, 2003, 11:22pm

[From Bruce Gregory (2003.12.07.1820)]

Bill Williams 7 December 2003 4:36 P.M.

Despite this he may, as you say, been controlling for the wrong
perceptions, but I wonder. In your opinion, if the pilot always
controls for the right perceptions, can the pilot be assured that
a PIO will never occur?

I don't know.

Bruce Gregory

"Everything that needs to be said has already been said. But since no
one was listening, everything must be said again."

Andre Gide

Bill_Williams1 · December 8, 2003, 12:14am

[Bill Williams 7 December 2003 6:14 P.M.]

Bruce,

Reading your reply to my question about PIO's it occurred to me that there may be PIO of more than one type.

The first and I think most familiar example is one in which there is excessive gain, or excessive loop delay resulting in oscilations. PIO are, I understand, quite common in air to air refueling.

But, on one occasion I observed a student at night become confused as to which direction was up. There was a full moon and broken clouds. By chance the full light of the moon fell on a lake which was highly reflective on a still night. Seeing the reflection of the moon below him on the lake, the student attempted to rotate the aircraft so that it would be, in his perception, upright. As he was doing this the student check the instruments and with his eyes off the out-side illusion corrected the planes attitude with reference to the instruments. Then the plane stablized in a genuinely upright position, the student went back to looking outside and repeating the cycle. This sequence repeated several times. Nothing wrong as far as I could tell with the students loop gain, or phase delay.

Bill Williams