Understanding

[Human_Factors_Lab]

[From Rick Marken (970223.0920)]

Bruce Abbott (970223.1145 EST)

_I've_ studied the Little Man demo, and understand how it works.
Does that make it three? (:->

Yes, indeed. At _least_ three!

But since you do understand the Little Man demo, I am more puzzled than
ever about why you say so many of the things of say in your
posts -- in particular, I don't understand the things you say about
model-based control.

Perhaps you could describe your understanding of the Little
Man demo. Maybe your understanding differs from mine.

Best

Rick

[Bruce_Abbott]

[From Bruce Abbott (970223.1545 EST)]

Rick Marken (970223.0920) --

But since you do understand the Little Man demo, I am more puzzled than
ever about why you say so many of the things of say in your
posts -- in particular, I don't understand the things you say about
model-based control.

I'm not surprised; I rather suspect that if you were to restate what you
_think_ I've said, I wouldn't even recognize it. (:-<

Perhaps you could describe your understanding of the Little
Man demo. Maybe your understanding differs from mine.

It is a hierarchically-organized feedback control system making no use of
inverse kinematics. It's been a while since I studied the code (over a year
now), but if I recall correctly there was a system to point the eyes at a
target by determining the horizontal and vertical visual angles of target
with respect to the line of sight and then adjusting the left-right and
up-down angle of the head as required to center the target in the eyes'
field of view (zero deviation from line of sight angles) via the usual
negative feedback mechanism.

Another set of control systems was used to set the vertical "finger tip" to
shoulder angle, the horizontal angle of the arm (pivoting at the shoulder),
and the distance between the shoulder and finger tip (as determined by the
elbow angle.) References for these three control systems were derived from
higher level systems controlling the difference between the visual angles of
the target and finger-tip as seen from the eyes (horizontal and vertical
deviations being used to set the references for the horizontal and vertical
finger-tip positioning systems, respectively, and the difference between
target and finger-tip distance from eyes (computed via parallax effects)
being used to set the reference for the lower-level finger-tip-to-shoulder
distance control system.

Because of this organization, as you moved the target around via the mouse,
the head turned to keep the eyes facing the target and the finger tip
followed, so that it continued to touch the target no matter where in the
permitted 3-D space you put the target.

That about it?

Understandingly,

Bruce

[Bill_Powers1]

[From Bill Powers (970223.1535 MST)]

Bruce Abbott (970223.1545 EST) --
(Replying to Rick Marken)
Bruce Gregory (970223) -- as an aid in studying the model.

It is a hierarchically-organized feedback control system making no use of
inverse kinematics. It's been a while since I studied the code (over a
year now), ...

Your memory is pretty good, but as is to be expected some of the details
tend to drop out. Allow me to fill some of them in. This is LM version 2 I
am talking about.

1. The eyes individually track the target, which maintains the image
disparity of the two target images at zero. The head moves 1/3 as much as
the eyes, not under control (but it could be). The image disparity for the
two images of the fingertip thus is a nonlinear measure of the radial
distance difference between fingertip and target. Actually, it's not
necessary to converge the eyes on the target; it's the _difference_ in image
disparities that is used to generate the visual depth signal. The reference
disparity-difference is zero, but can be set to any value. The x and y
position signals are measured relative to the center of the right retina.
The visual system controls the _difference_ in x, y, and z positions of the
finger and target, so there is no need to perceive positions in laboratory
space.

An internal signal-generator can be turned on which varies the x and y
target-to-finger distance reference signals in a sine and a cosine wave. The
result is that the finger traces a circle around the target, whether the
target is stationary or moving.

2. The outputs of the visual system in x, y, and depth set the reference
signals for a set of three second-level kinesthetic control systems that
also perceive fingertip position in x, y, and depth, as derived from
kinesthetic joint angle perceptions by simple means. The kinesthetic frame
of reference is centered on the shoulder joint. The depth signal is the
vertical shoulder angle minus half the elbow angle; it is nonlinear relative
to true physical distance. The x and y signals are simply the joint angles
corresponding to those directions. In the basic model there is no attempt to
make the kinesthetic coordinate system exactly match the visual one, or to
make it linear in objective space. This approach was taken to illustrate how
crude the control systems can be while still producing quite good control.

3. The second-level kinesthetic control system outputs set the reference
signals for the three first-level systems, which consist of models of the
stretch and tendon reflexes. The second-level depth or reach control system
actually contributes to the reference signals for both the shoulder vertical
angle and elbow angle systems at the first level, as well as perceiving a
variable made up of both related first-level perceptions. This illustrates
how one level of control can "coordinate" more than one lower-level control
system, in this case to produce approximately straight-line reaching
movements under control of a single reference signal.

4. Each first-level control system actually represents a pair of systems
operating in push-pull, with the calculations simplified to combine the
opposing effects. The muscle is modeled as a contractile element in series
with a spring element. The degree of contraction is proportional to the
error signal in each system; given a constant input signal, the muscle
produces a force that is proportional to the deviation of the joint angle
from a resting point set by the input signal. The muscle contraction is
modeled as a leaky integrator to reproduce, roughly, the 50-millisecond time
constant of contraction (the time-constant of relaxation is assumed to be
the same, contrary to fact). If the joint is clamped in a fixed position,
varying the driving signal will produce a torque at the joint as the series
spring element is stretched. If the joint is free to move, the joint angle
will accelerate toward the current resting position set by the driving signal.

5. The joint angles are directly linked to muscle length. Muscle length is
sensed by models of the spindle cells in that there is a tonic and a phasic
component of the signal, and a gamma-efferent signal sets the length
reference level directly in the muscle spindle (a mechanical comparator).
The stretch signal is thus actually an error signal, carrying rate plus
proportional error information to the spinal motor neuron, which is the
comparator for the tendon reflex control system. It is assumed that the
alpha and gamma efferent signals are co-activated.

6. The tendon signal is affected by the force generated in the muscle, and
is proportional to the stretch in the series spring component. It feeds back
(negatively) directly to the spinal motor neuron. It is a direct measure of
the torque applied at the joint.

7. The three torques which are the outputs of the combined stretch and
tendon control systems cause angular accelerations around the three joints.
These torques are the inputs to a (forward) physical model of the
three-degree-of-freedom arm. The physical model of the arm converts these
torques into angular accelerations, including all interactions among the arm
segments such as Coriolis forces. The angular accelerations, integrated
twice, return values of the three joint angles updated once per iteration.
The joint angles are simply limited at the physiological limits, as measured
with my own arm. These joint angles determine the lengths of the muscles and
thus affect the spinal reflex systems, and so on back up through the stack
of control systems. From the joint angles and the dimensions of the arm
segment, we also calculate the position of the fingertip in space (really,
the end of the second arm segment). From that position and the look-angle of
the eyes, the position of the image of the fingertip on each retina is
calculated.

Paragraph 7 is thus a description of the environment of the control systems
with respect to the mechanical and visual variables of importance. The
effect of gravity is also included, and gravity can be switched on and off
while the model runs.

The environmental model, which includes arm dynamics, and the model of the
reflexes, run four iterations for each iteration of the higher levels of
control. If the time per iteration at the higher levels is taken to be 30
per second, then the time per iteration at the lowest level is 120 per
second, or about 8 milliseconds each. This is as close as the model comes to
dealing with time-delays at the different levels.

There are test modes for the first and second kinesthetic levels, in which
square-wave reference signals are used to exercise each degree of freedom
while the user adjusts the parameters of the control systems. The entire
system can be tested by moving the target around, either manually, back and
forth between fixed end-points, or in random jumps.

A "mapping" mode exists in which the relation between the visual and
kinthetic systems (which have different origins) is gradually adjusted. It
works very poorly and I don't recommend using it. That's just where I was
when I ran out of energy in developing this model, and it's probably not a
good idea.

The visual level of control works reasonably well, but needs a lot of work
to optimize it. In fact the whole model could use a lot of work, introducing
better muscle models and representing the opposing muscle systems
explicitly. There is lots of information available in the literature of
physiology, and I had hoped to built a Little Man Version 3 that would bring
it into the model. However, even as it stands, the model is a clear
demonstration of the simplicity of a feedback hierarchy as an explanation of
pointing behavior, and as a direct demonstration that feedback is NOT too
slow. I have tried to get people like Bizzi and Gomi to look at it, but in vain.

Best,

Bill P.

[Bruce_Abbott]

[From Bruce Abbott (970223.2130 EST)]

Bill Powers (970223.1535 MST) --

Your memory is pretty good, but as is to be expected some of the details
tend to drop out. Allow me to fill some of them in. This is LM version 2 I
am talking about.

No doubt I slipped on a few details, but not THAT much! The version I'm
familiar with (and was describing) is LM version 1. I gather from your
description of LM 2 that this version adds the details of the lower-level
muscle systems that were only assumed in LM 1. Last year, as you know, I
did a little work on the "Isaac" arm model, which modeled the physics of a
forearm, jointed at the elbow, whose angle with respect to the horizontal is
determined by setting a reference for joint angle; this joint angle control
system's output set the reference for the lower-level muscle-length control
system. As in LM 2, I treated the arm as a damped spring; it was quite an
interesting exercise to implement the physics correctly in iterated computer
code.

Regards,

Bruce

[Bill_Powers1]

[From Bill Powers (970224.0830 MST)]

Bruce Abbott (970223.2130 EST) --

No doubt I slipped on a few details, but not THAT much! The version I'm
familiar with (and was describing) is LM version 1. I gather from your
description of LM 2 that this version adds the details of the lower-level
muscle systems that were only assumed in LM 1.

Yes, that addition came along in about 1989 or 90 (I haven't been good about
preserving versions, so I don't know exactly when the first working version
was done). I do believe the source code is on the Web page.

Last year, as you know, I
did a little work on the "Isaac" arm model, which modeled the physics of a
forearm, jointed at the elbow, whose angle with respect to the horizontal
is determined by setting a reference for joint angle; this joint angle
control system's output set the reference for the lower-level
muscle-length control system. As in LM 2, I treated the arm as a damped
spring; it was quite an interesting exercise to implement the physics
correctly in iterated computer code.

I don't recall the details of your model, but remember that it worked.
Actually, in LM2, the muscle is a mass on a spring, but the driving signal
changes the resting length of the spring rather than simply producing an
output force; that's closer to the actual mode of operation of a muscle. The
forward dynamics of the arm is a good deal more complex than that of a mass
swinging about a single fixed pivot; the arm has three degrees of freedom
(pitch and yaw at the shoulder, pitch at the elbow), and there are
interactions among these degrees of freedom (as expressed in the dynamical
model). For example, suddenly straightening the elbow exerts a torque
tending to raise the upper part of the arm, and raising the arm from the
shoulder tends to straighten the elbow if the interior elbow angle is
greater than 90 degrees, but tends to fold it up if that angle is less than
90 degrees. Also, swinging the partly-extended arm sideways creates a
centrifugal force tending to straighten the arm, and retracting the arm
while it is swinging sideways tends to speed up the swing (Coriolis force).
The neat thing about the model is that it works without any elaborate
attempts to compensate for these interactions among the independent control
systems. There's only one built-in compensation, and it's real: some of the
biceps-triceps force is applied to both raise the upper arm and bend the
elbow: this represents a rough model of the part of these muscles that spans
both the shoulder and elbow joints (two-joint attachments). The dynamical
and kinematic compensation falls out of the models for the individual
systems. Each system simply treats effects from the other systems as
disturbances, and resists them.

The damping in this model comes mostly from the phasic component of the
stretch reflex signal. I believe that this model addresses every question,
in at least a preliminary proof-of-principle way, that the "forward control"
modelers have considered, if not more.

With regard to the eye model, the "retinal position" is simply the
difference between the look angle to the target or fingertip (two
dimensions) and the optic axis of the eye, and is calculated as an angle.
However, multiplying this angle by the diameter of the eye would give the
retinal position relative to the fovea. I didn't bother; that would only
affect a scaling factor.

···

----------------------------------------
Avery Andrews (970224) --

The journal is just called "Neural Networks," ISSN 0893-6080-841, published
under Pergamon Press. The Forward Models article is in Vol 9, No 8, on pp.
1265-1279 (1996).

Best to all,

Bill P.

[Bruce_Gregory6]

[From Bruce Gregory (2005.03.26.2005)]

Ely Dorsey wrote:

The philosophy is very rich. General Relativity also helps one see
sides of QT that are different. This theory is not hard at all. Anyone
can play with it. The difficult part is the nature of explanation. That
is, what do we mean by explanation? This is why I am studying your work
along with Maturana and von Glasersfeld.

The most useful comment I have found in this regard comes from John von Neumann,

"In mathematics you don't understand things, you just get used to them."

This is even more true of physics, I have found.

A true believer knows the solution before he understands the problem.

[Bill_Powers1]

[From Bill Powers (2005.03.26.0840 MST)]

Bruce Gregory (2005.03.26.2005)--

The most useful comment I have found in this regard comes from John von Neumann,

"In mathematics you don't understand things, you just get used to them."

This is even more true of physics, I have found.

In some ways, yes. But there is usually more than one level of explanation available, even when you think you've reached the bottom level. Look at so-called "elementary" particles, which eventually were explained as compounds of quarks.The quarks are the mechanisms that explain how particles combine and break up. Now, as I understand it, people are asking what mechanism keeps quarks so tightly bound together. It probably never ends. Some day people will be asking what strings are made of.

In PCT, the "elementary" level is that of the neural circuit, a computing device that receives neural signals (or stimuli) as inputs and produces new neural signals that are functions of those inputs. This provides an explanation of how some neural signals depend on others and on external variables. Of course one can then dig deeper and explain the neural computing functions in terms of biochemical mechanisms inside neurons. How far down you go depends on what you're trying to explain, because there are orderly phenomena unique to every level. If you're trying to explain experience and behavior, there's no need to go into biochemistry; that would be like explaining how a radio works in terms of electrons and holes in semiconductors. The farther down you go, the more details there are, and they quickly become incomprehensible when considered in the context of global phenomena. Also, at any given level there is more than one way to combine lower-level mechanisms to achieve any given effect at a higher level; much of the detail revealed by going down a level is irrelevant to the phenomenon initially considered. You can make a chair in a zillion ways, but the result is still something to sit on.

Best,

Bill P.

···

A true believer knows the solution before he understands the problem. </blockquote></x-html>

[Ely_Dorsey]

Hi Bruce,

Thanks for joining the discussion. There is a movie
out on Quantum Theory, “What the Bleep Do You Think You Know?!!”
You can but it at Blockbuster. It is an amazing film. You can see
it twenty times and see something different each time. It is worth the
buy.

In the Rae book, he speaks about the need for QT to also
explain Classical physics. It is all very curious, this idea of
explanation. Bill made it quite clear that the idea of causality had to
be seriously questioned when one views systems from a CT perspective.
What struck me was his insight into the non autonomous nature of the reference
signal. It appeared to me that the whole system could be viewed this way.
QT also questions causality in a similar way. Well, I am getting ahead of
myself.

If you look at the Mermin or Aspect experiments attached to
all of this, you see this unexplained correlation of photons occurring.
Well, I was thinking that the correlation itself was a reference signal of
sorts to the system of the experiment. It is a just a way to talk about
this.

I want to see if we can speak of these quantum experiments
in terms of CT. If so, what would the discussion be like. I think Bill’s
work lends itself to this different dialogue.

Ely

···

---

From: Control Systems
Group Network (CSGnet) [mailto:CSGNET@listserv.uiuc.edu] On Behalf Of Bruce Gregory
Sent: Saturday, March 26, 2005
8:55 AM
To: CSGNET@listserv.uiuc.edu
Subject: Understanding

/x-tad-smaller>/fontfamily>[From Bruce Gregory (2005.03.26.2005)]

Ely Dorsey wrote:

The philosophy is very rich. General Relativity also helps one see

sides of QT that are different. This theory is not hard at all. Anyone

can play with it. The difficult part is the nature of explanation. That

is, what do we mean by explanation? This is why I am studying your work

along with Maturana and von Glasersfeld.

/x-tad-smaller>/fontfamily>The most useful comment I have found in this
regard comes from John von Neumann,

“In mathematics you don’t understand things, you just get used to
them.”

This is even more true of physics, I have found.

A true believer knows the solution before he understands the problem.

[Ely_Dorsey]

HI Bill and Bruce,

In QT there is a measurement problem: you come up against a limit of size
and energy when attempting to measure subatomic things. The Heisenberg
Uncertainty Principle comes into play as well as the physicality of the
measurement apparatus. There is a school of thought, which you will read
about in the Rae book, it is called the Copenhagen View.

"What is now the orthodox approach to quantum physics adopts this radical
point of view, questioning whether the postulate of some correlation between
the state of a measuring apparatus and that of a distant photon is
meaningful and indeed whether photons can be said to have any existence at
all until they are in s some sense observed." (Rae, p.51)

What happens in QT is that the measurement and observation of an observable
begins a dance into the realms of philosophy, but the classical mind demands
that the music stop and an accounting occur. I think Bills work will help
classical mind see the dance.

···

-----Original Message-----
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@listserv.uiuc.edu] On Behalf Of Bill Powers
Sent: Saturday, March 26, 2005 10:59 AM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Understanding

[From Bill Powers (2005.03.26.0840 MST)]

Bruce Gregory (2005.03.26.2005)--

The most useful comment I have found in this regard comes from John von
Neumann,

"In mathematics you don't understand things, you just get used to them."

This is even more true of physics, I have found.

In some ways, yes. But there is usually more than one level of explanation
available, even when you think you've reached the bottom level. Look at
so-called "elementary" particles, which eventually were explained as
compounds of quarks.The quarks are the mechanisms that explain how
particles combine and break up. Now, as I understand it, people are asking
what mechanism keeps quarks so tightly bound together. It probably never
ends. Some day people will be asking what strings are made of.

In PCT, the "elementary" level is that of the neural circuit, a computing
device that receives neural signals (or stimuli) as inputs and produces new
neural signals that are functions of those inputs. This provides an
explanation of how some neural signals depend on others and on external
variables. Of course one can then dig deeper and explain the neural
computing functions in terms of biochemical mechanisms inside neurons. How
far down you go depends on what you're trying to explain, because there are
orderly phenomena unique to every level. If you're trying to explain
experience and behavior, there's no need to go into biochemistry; that
would be like explaining how a radio works in terms of electrons and holes
in semiconductors. The farther down you go, the more details there are, and
they quickly become incomprehensible when considered in the context of
global phenomena. Also, at any given level there is more than one way to
combine lower-level mechanisms to achieve any given effect at a higher
level; much of the detail revealed by going down a level is irrelevant to
the phenomenon initially considered. You can make a chair in a zillion
ways, but the result is still something to sit on.

Best,

Bill P.

A true believer knows the solution before he understands the problem.
</blockquote></x-html>

[Bruce_Gregory6]

[From Bruce gregory (2005.0326.1347)]

Bill Powers (2005.03.26.0840 MST)

Actually, I was thinking of something simpler. I've come to the conclusion that we can't really understand Newton's Laws. When a baseball is hit into the air, it makes sense that the force exerted by the mass of the earth pulls the baseball back to the ground. I suggest that it does not make sense that the tiny baseball exerts exactly the same force on the earth that the earth exerts on the baseball. How is that possible? How can a baseball exert as much force as a planet?

So I claim you never really understand Newton's Third Law. You just get used to it.

A true believer knows the solution before he understands the problem.

[Bozo_the_Clown]

Well, I was thinking that the correlation itself was a reference signal of sorts to the system of the experiment. It is a just a way to talk about this.

I want to see if we can speak of these quantum experiments in terms of CT. If so, what would the discussion be like. I think Bill’s work lends itself to this different dialogue.

Ely, you bring up some extremely important points about doing science here that have fallen on deaf ears here on CSGnet. Maybe they will listen to you.

There are many ways to think about things, explain things, and to view them. This is one reason why science, if done properly, is never done as a matter of being ‘right’ or ‘wrong’. Since certainty is impossible, all we can ever hope to accomplish is to strive for some approximation of the truth.

The notion Bill has that PCT or some other theory will prove to be ‘right’ or ‘wrong’ is bad science and misguided because it cuts off the dialogue among people. If you are convinced someone is ‘wrong’ there is never a need to listen to anything the person has to say, convinced that someone else’s view is worthless just eliminates a potential useful source of information.

When ideology take a back seat exploration, science ceases to exist and when you stop exploring, the search for truth stops as well.

I’m convinced Bill Powers feels he has already found the ‘truth’ in PCT and most of the folks here believe that as well.

You mentioned a few folks that have been involved in the development of systems ideas. How familiar are you with the control community? Are you familiar with System Dynamics and the work of Jay Forrester? You might want to look into that and a great deal more as well in the field of control. You might also want to confer with Cliff Joslyn on this. I would also hiighly recommend Feedback Thought i the Social Sciences by George Richardson as a way for you to take an historical view of the field.

All of this ‘advice’ of course predicated on the notion that science and not religion is your main pursuit.

Marc

···

Ely

---

From: Control Systems Group Network (CSGnet) [mailto:CSGNET@listserv.uiuc.edu] On Behalf Of Bruce Gregory
Sent: Saturday, March 26, 2005 8:55 AM
To:
CSGNET@listserv.uiuc.edu
Subject:
Understanding

/x-tad-smaller>/fontfamily>[From Bruce Gregory (2005.03.26.2005)]

Ely Dorsey wrote:

The philosophy is very rich. General Relativity also helps one see
sides of QT that are different. This theory is not hard at all. Anyone
can play with it. The difficult part is the nature of explanation. That
is, what do we mean by explanation? This is why I am studying your work
along with Maturana and von Glasersfeld.

/x-tad-smaller>/fontfamily>The most useful comment I have found in this regard comes from John von Neumann,

“In mathematics you don’t understand things, you just get used to them.”

This is even more true of physics, I have found.

A true believer knows the solution before he understands the problem.

[Bill_Powers1]

[From Bill Powers (2005.03.27.0705 MST)]

Ely Dorsey [2005.03.26) –

Bill made it quite clear that the idea of causality had to be
seriously questioned when one views systems from a CT perspective.
What struck me was his insight into the non autonomous nature of the
reference signal. It appeared to me that the whole system could be
viewed this way. QT also questions causality in a similar
way.

It is not the idea of causality that has to be questioned in control
theory, but the particular version of that idea that cause-effect or S-R
psychology has assumed. In a control system, each element of the system
is a normal cause-effect process which converts inputs into outputs, but
the assembly of elements is wrapped into a closed loop. This means that
when you trace the cause of any variation in the loop to its antecedents,
eventually you discover that one of the antecedents was a previous state
of that same variable. So in terms of ultimate causation, every
element of the closed loop causes itself, or is one of the causes of its
own future states.

I wouldn’t say, therefore, that control theory deals with causation
“in the same way” that quantum theory does. In quantum theory,
events cause other events simply by virtue of happening, which is
certainly not the case in control theory. Also, in quantum theory an
event takes place at an instant of time rather than being a process that
occurs through time. There is no such thing as being halfway through a
transition from one quantum state to another. That, also, is very
different from control theory, in which all variables are assumed to be
continuous and extended in time in their basic nature (because they are
represented in the nervous system by the frequency of trains of impulses,
and frequency is not even defined at one instant).

Perhaps when I read the Rae book some of my problems with QT will be
resolved.

If you look at the Mermin or Aspect experiments attached to all of this,
you see this unexplained correlation of photons occurring. Well, I
was thinking that the correlation itself was a reference signal of sorts
to the system of the experiment. It is a just a way to talk about
this.

A reference signal is a variable that enters into a control process in a
very specific way, but otherwise it is just an ordinary physical
variable, a train of neural impulses in PCT. I don’t see how a
correlation, which is a computation of relatedness, shares any
characteristics or any functions with a reference signal. If you have in
mind that two variables SHOULD correlate to some specific degree, then I
suppose that degree of relatedness could become a reference signal
(“I want to perceive a better correlation between your predictions
and what actually happens, Mr. Jones”). Then the actual degree of
correlation would have to be continuously or repeatedly calculated to
provide a perceptual signal, and the two variables, reference and
perception, could be compared to yield an error signal which would drive
some activity that affects future values of the perceived correlation.
But I don’t see how that interpretation would relate to quantum
theory.
If you have to say “a reference signal of sorts” then
clearly you’re speaking metaphorically and not literally. What exactly is
it that you see as being the same in a reference signal as it is in a
correlation?

Best,

Bill P.

[Bill_Powers1]

[From Bill Powers (2005.03.27.0730 MST)]

Ely Dorsey (2005.03.26b)

It really would be easier to refer to your posts if you began with the standard date-time stamp we use on CSGnet. The format is

[From <name here> (yyyy.mm.dd.tttt)] (see first line of text above).

When replying, you simply delete the "[From" and the closing bracket, and start your reply on the next line. This way you see who the post is from and when it was composed right away, and don't have to search for the time or date elsewhere.

Don't forget to delete the parts of the post that are not relevant to your comments. Too manmy posts consist of "Yeah, that's a good point" followed by three pages of copied text in which the supposed point is concealed somewhere.

···

===============================

In QT there is a measurement problem: you come up against a limit of size
and energy when attempting to measure subatomic things. The Heisenberg
Uncertainty Principle comes into play as well as the physicality of the
measurement apparatus. There is a school of thought, which you will read
about in the Rae book, it is called the Copenhagen View.

I always thought that the Copenhagen view went deeper than mere uncertainty. It claimed specifically that the uncertainty was in the phenomenon rather than in our measurement of it. In short, measuring position and momentum was not uncertain just because our measurements perturbed one or both quantities, but because they were inherently indeterminate on the scale where effects of Planck's Constant became discernible.

At one point in my career, someone explained to me that this was basically a philosophical position -- that physics was a science based on observations, not direct knowledge of reality, so we could not say anything about reality itself. All we could talk about was observations. Thus uncertainty in observations was, in a sense, uncertainty in reality, but not because there was any uncertainty in the natural world. It was so because reality itself, to a physicist, exists only in the form of observations and calculations.

I accepted this as the proper interpretation of the Copenhagen view, even though I later encountered many physicists who interpreted it as the naive realist would: the uncertainty exists in the natural world, apart from us observers of it. I think that this is why I have never been able to swallow the explanations offered by quantum physicists (Feynman aside, because he admitted that he simply made it all up).

It's interesting that only now do I see the connection between what that unknown physicist said to me about the Uncertainty Principle and my current views on PCT. It is, indeed, all perception, including the apparent uncertainties we observe. I also realize that another thing stuck in my craw: a physicist could claim that physics was a science of observation, while at the same time ignoring the fact that he himself was observing and interpreting what his artificial instruments seemed to be telling him. Physics has presented itself as the ultimate science because it deals directly with nature through objective and systematic experimentation. But physicists have never liked to be reminded of the eyes with which they see their instruments, the hands with which they manipulate the experiments, and the perceptions of the results in their brains which are all they can ever know of nature. That sort of stuff is handled over on the other end of the campus where the fuzzies live. Physics, as a system concept, has attained a kind of independent reality, and is viewed as something that exists in the air between physicists rather than in the signals racing around in their brains. Physicists, particularly quantum physicists, seem to have lost sight of their own human roles as the inventors and interpreters of physics.

The link between QT and PCT that I see, therefore, is of the nature of what Alexander Turchin calls a "metasystem transition." PCT is based in part on recognition that all human activities, including physics, are control processes perceived and controlled in and by human brains. As in the case of all human interactions with the natural world, part of what we experience is due to properties of the world outside, but those properties are known to us only in the form of neural computations in a brain, and are therefore influenced heavily by the nature of the brain. How much of what we know about the external world is actually an aspect of the way our brains work? We will never know until we encounter another living system which has evolved in total isolation from Earthly conditions.

I will be eager to see what Rae says about such things.

Best,

Bill P.

[Ely_Dorsey]

From Ely Dorsey 2005.03.27.14:23 EST

Hi Bill,

I am going to answer somewhat both of your notes. The Copenhagen view is
that reality comes into being only after observation. So the act of
observation is a signal. The question is a signal to what? I am asking you
to view the quantum experiments the way a control theorist would. Where is
the Comparator, what are the inputs etc. Does a quantum experiment seek
stability? These are the type of questions I am putting to you. Avoid
being too precise with language for right now; instead think model and
levels in the most open sense. We do not have to be right or wrong at this
moment, just imaginative. We want to come out of this with a control
theorist's view of quantum theory. To do this, we have to let go of what we
know and be inside what we want to know and then back out slowly like
walking through a Klein bottle.

For example, since the photons are correlated every time they are detected
in the same position, then can we consider every time they are detected in
different positions, errors?

Another example, since we cannot measure simultaneously the momentum and
position of a subatomic particle, is this an error or a reflection of a
reference signal that is varying?

What I am getting at Bill, is that QT is ordered, but not in the way we
understand order. CT could shed light on that notion of ordering. For
this, we have to create thought experiments in CT.

Ely

···

-----Original Message-----
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@listserv.uiuc.edu] On Behalf Of Bill Powers
Sent: Sunday, March 27, 2005 10:12 AM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Understanding

[From Bill Powers (2005.03.27.0730 MST)]

===============================

In QT there is a measurement problem: you come up against a limit of size
and energy when attempting to measure subatomic things. The Heisenberg
Uncertainty Principle comes into play as well as the physicality of the
measurement apparatus. There is a school of thought, which you will read
about in the Rae book, it is called the Copenhagen View.

I always thought that the Copenhagen view went deeper than mere
uncertainty. It claimed specifically that the uncertainty was in the
phenomenon rather than in our measurement of it. In short, measuring
position and momentum was not uncertain just because our measurements
perturbed one or both quantities, but because they were inherently
indeterminate on the scale where effects of Planck's Constant became
discernible.

At one point in my career, someone explained to me that this was basically
a philosophical position -- that physics was a science based on
observations, not direct knowledge of reality, so we could not say anything
about reality itself. All we could talk about was observations. Thus
uncertainty in observations was, in a sense, uncertainty in reality, but
not because there was any uncertainty in the natural world. It was so
because reality itself, to a physicist, exists only in the form of
observations and calculations.

I accepted this as the proper interpretation of the Copenhagen view, even
though I later encountered many physicists who interpreted it as the naive
realist would: the uncertainty exists in the natural world, apart from us
observers of it. I think that this is why I have never been able to swallow
the explanations offered by quantum physicists (Feynman aside, because he
admitted that he simply made it all up).

It's interesting that only now do I see the connection between what that
unknown physicist said to me about the Uncertainty Principle and my current
views on PCT. It is, indeed, all perception, including the apparent
uncertainties we observe. I also realize that another thing stuck in my
craw: a physicist could claim that physics was a science of observation,
while at the same time ignoring the fact that he himself was observing and
interpreting what his artificial instruments seemed to be telling him.
Physics has presented itself as the ultimate science because it deals
directly with nature through objective and systematic experimentation. But
physicists have never liked to be reminded of the eyes with which they see
their instruments, the hands with which they manipulate the experiments,
and the perceptions of the results in their brains which are all they can
ever know of nature. That sort of stuff is handled over on the other end of
the campus where the fuzzies live. Physics, as a system concept, has
attained a kind of independent reality, and is viewed as something that
exists in the air between physicists rather than in the signals racing
around in their brains. Physicists, particularly quantum physicists, seem
to have lost sight of their own human roles as the inventors and
interpreters of physics.

The link between QT and PCT that I see, therefore, is of the nature of what
Alexander Turchin calls a "metasystem transition." PCT is based in part on
recognition that all human activities, including physics, are control
processes perceived and controlled in and by human brains. As in the case
of all human interactions with the natural world, part of what we
experience is due to properties of the world outside, but those properties
are known to us only in the form of neural computations in a brain, and are
therefore influenced heavily by the nature of the brain. How much of what
we know about the external world is actually an aspect of the way our
brains work? We will never know until we encounter another living system
which has evolved in total isolation from Earthly conditions.

I will be eager to see what Rae says about such things.

Best,

Bill P.

[Bozo_the_Clown]

Ely,

Are you addressing these questions and posts specifically to Bill or are they open for all to participate in?

Marc

[Dick_Robertson]

[FRom Dick Robertson, 2005.03.28.1330CDT]

Enjoying the discussion of understanding and QT. I could offer a little
conjecture on one of your points (BP) in the following.

···

----- Original Message -----
From: Bill Powers <powers_w@FRONTIER.NET>
Date: Sunday, March 27, 2005 9:11 am
Subject: Re: Understanding

==============

>In QT there is a measurement problem: you come up against a limit
of size
>and energy when attempting to measure subatomic things. The
>Uncertainty Principle comes into play as well as the
physicality of the
>measurement apparatus. There is a school of thought, which you
will read
>about in the Rae book, it is called the Copenhagen View.

I always thought that the Copenhagen view went deeper than mere
uncertainty. It claimed specifically that the uncertainty was in
the
phenomenon rather than in our measurement of it. In short,
measuring
position and momentum was not uncertain just because our
measurements
perturbed one or both quantities, but because they were inherently
indeterminate on the scale where effects of Planck's Constant
became
discernible....

It's interesting that only now do I see the connection between what
that
unknown physicist said to me about the Uncertainty Principle and my
current
views on PCT. It is, indeed, all perception, including the apparent
uncertainties we observe. I also realize that another thing stuck
in my
craw: a physicist could claim that physics was a science of
observation,
while at the same time ignoring the fact that he himself was
observing and
interpreting what his artificial instruments seemed to be telling
him.
Physics has presented itself as the ultimate science because it
deals
directly with nature through objective and systematic
experimentation. But
physicists have never liked to be reminded of the eyes with which
they see
their instruments, the hands with which they manipulate the
experiments,
and the perceptions of the results in their brains which are all
they can
ever know of nature. That sort of stuff is handled over on the
other end of
the campus where the fuzzies live. Physics, as a system concept,
has
attained a kind of independent reality, and is viewed as something
that
exists in the air between physicists rather than in the signals
racing
around in their brains. Physicists, particularly quantum
physicists, seem
to have lost sight of their own human roles as the inventors and
interpreters of physics.

I propose that this is a characteristic not simply of physicists but of
"science" in general since Descartes. What I call the Cartesian
psychology has so infused all thinking about behavior, any kind
including scientific, that it never occurs to the scienctist, or any
other citizen that we could do other than to look at phenomena from the
outside in, (because the "laws" of science, whether physical or
psychological are about the nature of "reality," which we stand outside,
observing.

The link between QT and PCT that I see, therefore, is of the nature
of what
Alexander Turchin calls a "metasystem transition." PCT is based in
part on
recognition that all human activities, including physics, are
control
processes perceived and controlled in and by human brains. As in
the case
of all human interactions with the natural world, part of what we
experience is due to properties of the world outside, but those
properties
are known to us only in the form of neural computations in a brain,
and are
therefore influenced heavily by the nature of the brain. How much
of what
we know about the external world is actually an aspect of the way
our
brains work? We will never know until we encounter another living
system
which has evolved in total isolation from Earthly conditions.

I will be eager to see what Rae says about such things.

Best,

Dick R

[Bill_Powers1]

[From Bill Powers (2005.03.28.1237 MST)]

Ely Dorsey 2005.03.27.14:23 EST --

I am going to answer somewhat both of your notes. The Copenhagen view is
that reality comes into being only after observation. So the act of
observation is a signal. The question is a signal to what?

I should think, to the observer. Of course there is a reciprocal effect on whatever is observed, but it is small and anyway doesn't differ when the observer draws different conclusions about the observed.

I am asking you to view the quantum experiments the way a control theorist would. Where is the Comparator, what are the inputs etc. Does a quantum experiment seek stability?

You can't talk about the experiment without mentioning the experimenter. It's the experimenter who has perceptions, reference signals, error signals, and so on, not the experimental apparatus or the phenomenon being observed and affected. I'm having a hard time imagining what you mean here. How can an experiment seek anything? It's just a procedure.

These are the type of questions I am putting to you. Avoid
being too precise with language for right now; instead think model and
levels in the most open sense.

I don't really want to get too sloppy about this. I know we can't always be precise when using a tool as imprecise as language, but if we start letting the meanings of words like reference and error drift all over the place we'll be losing ground, not gaining it.

We do not have to be right or wrong at this moment, just imaginative. We want to come out of this with a control theorist's view of quantum theory.

I think the best we can hope for is to come out of this with a control theorist's view of quantum theorists. PCT is about people. Unless you mean that you want to look for negative feedback explanations of quantum phenomena.

To answer someone else's comment, Perceptual Control Theory is just Negative Feedback Control Theory. All negative feedback system control the input affected by their own actions, so they all control perceptions rather than outputs. You can view inertia as a negative feedback effect if you stick your neck out. Are you thinking of something similar for quantum phenomena?

For example, since the photons are correlated every time they are detected
in the same position, then can we consider every time they are detected in
different positions, errors?

Whose error do you mean? An error in PCT is a difference between what is perceived and what is intended to be perceived. Do photons intend to perceive anything? Do they perceive anything? I think an experimenter can intend to perceive something, and experiences errors when the actual perception differs from what was intended. Are you proposing that photons are also control systems?

Another example, since we cannot measure simultaneously the momentum and
position of a subatomic particle, is this an error or a reflection of a
reference signal that is varying?

Again you have to specify the system in which the supposed error and reference signal are supposed to exist. Are you proposing that they exist in the subatomic particle? I assume they exist only in the observer, the person doing the measurement.

What I am getting at Bill, is that QT is ordered, but not in the way we
understand order. CT could shed light on that notion of ordering. For
this, we have to create thought experiments in CT.

"Ordering" by itself doesn't mean anything. There must be something to be ordered, and there must be some regular principle of ordering. What is ordered about QT?

My current impression of QT is that it consists of a lot of explanations that don't yet hold water. The phenomena are of great interest and are very puzzling, but so far I think we have only some pretty bad guesses as to what is really going on. When we come up with a good guess, everyone will say, "Oh, of course, now it all makes sense."

When I read Rae I may have to change my mind.

Best,

Bill P.

[Marken_Richard_S1]

[ From Rick Marken (930614.1400)]

Tom Hancock (930614)--

In a PCT view, we could say
that understanding is when the perception is made to match its
reference signal.

My first definition of "understanding" was that it is "knowledge" embodied
in the comparator and output functions that makes it possible for the
control system to match its perceptual to its reference signal. What you
describe above already has a name, I think -- it's called "control".

when I consider students in my classes,
who are perceiving something when I speak or demonstrate, what
could be happening when they have 100% confidence of understanding
("I get it perfectly") versus when they have 50% ("I sort of
understand") or 0% (I am totally confused).

There are many possible answers to this question, many of which are
more questions. I presume you have asked the students to rate their
confidence that they have "understood" you. What does each of the
students think he or she is rating? Each probably has a different idea
of "understanding". Maybe they are trying to give a number that is an
estimate of what their score on a test of the material might be?
Maybe they are trying to impress you-- or win your sympathy. Or maybe
they have no clue what you are talking about except that you want them
to say a number; and they say it.

I think the problem may be that "understanding" is just a word; what is the
phenomenon to which it refers, for you?

Are they controlling for a
perception of an "integration" of prior perceptions into a
distinct perception (a partially "cognitive" view)?

Maybe. That would be a tough one to test until it were made more
concrete. How can I observe the degree of integration of prior
perceptions into a distinct perception? If this is what they are
controlling, it is likely to be something that you can perceive
as well. It should also be a perception that can vary and, if
it is controlled, is maintained in a reference state against
disturbance. I've never had a perception that I would
decribe as "integration of prior perceptions into a distinct
perception". Maybe there's a better way to describe it.

In a general sense, would you say that degrees of understanding is
simply indicative of unsuccessful attempts to bring perceptions
into line with references, and as such would the degree of
understanding be inversely related to the magnitude of some
persisting error signals?

No. I think what you are describing is just "poor control", not lack
of understanding.

Rick, could you help out here and articulate a more specific HPCT
example of what a model demonstrating degrees of understanding
would look like with an individual who is not involved in a
physical procedural task such as throwing a baseball or threading
a needle, nor is he explaining anything, but is just sitting in a
chair, listening and "trying to understand" the teacher.

I think you have to get a better handle on the phenomenon that
you are trying to study. My guess is that you are interested in
what happens when a person get's some verbal communications
about how to do something (such as build a radio or solve algebra
equations but it could also include giving the right answers on a test).
A person "understands" to the degree that they can create, in their
imagination, an image of themselves succeeding at the thing that they
are supposed to do after the lecture. So, in your class, I presume most
of the students are trying to produce lots of correct answers on your
test. As you talk they are trying to memorize what you say, maybe
determine what you might ask on the test, how you might ask it, whether
their replies would satisfy you (so they ask questions or volunteer
information as a test what you might consider to be "right" answers).
So they are controlling, in imagination, for how well they can do on one
of your tests. This is just a bunch of guesses, but some might be
testable (using The Test).

Maybe you could try explaining once again what you are trying to
understand. What have you observed that is interesting to you and
that has motivated your research? What have you seen that you
want to explain? If you have seen "understanding" then please
explain in concrete terms what it is that you've seen.

I think you are interested in a very interesting phenomenon. I
just think that a big part of the problem of studying it in PCT
terms is identifying the perceptual variables that are involved
(both for you and for the people you are studying) in this phenomenon
that you call "understanding".

Best

Rick

[Abbott_Bruce]

[From Bruce Abbott (991115.1330 EST)]

Apparently, the ability to understand either Bill or me is
decreased considerably if your first name happens to be Bruce.
Now _that_ is a puzzling phenomenon;-)

Nothing too puzzling about it -- all the really smart people are named Bruce.

Regards,

Bruce