Testing for Controlled Variables (Re: Feedback Functions)

[From Rick Marken (2015.02.15.1510)]

···
MT: Do you intend to imply that the behaviour of a control system is NOT

what is predicted by causal law?

RM: No, it’s the behavior of the hypothetical controlled variable in response to disturbance that is not predicted by causal law when the variable is, indeed, controlled. For example, if we want to know whether or not the position of a mass suspended by a spring is under control we test this by applying a force disturbance, Fd, to the mass to see if it has the expected effect on the position of the mass – expected from causal law.

RM: The causal law that gives the predicted effect of the force on the position of the mass is Hooke’s law. The predicted effect of the disturbance is a change in the position of the mass, x. The predicted size of x in response to force disturbance Fd, is x = 1/s*Fd. So if we know the spring constant, s, and Fd, then we predict that applying this force to the mass will result in x amount of change in the position of the mass.

RM: If the position of the mass is not under control then the observed change in position of the mass, x’, when force Fd is applied will exactly equal x, per the causal laws of physics (approximated in Hooke’s law). If x’<<x then the position of the mass is likely under control; it is a controlled variable.

RM: This is the test for the controlled variable (TCV). Basically you are looking to see if the observed effect of a disturbance to a hypothetical controlled variable is less than expected from causal law. An understanding of causal law is essential to being able to do the TCV.

RM: Counting the restoring force exerted by the spring when Fd is applied as resistance to disturbance just confuses things because that language implies that this resistance is preventing the disturbance from having an effect, and of course it’s not.

Best

Rick


Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

        RM: As I said to Bruce in an earlier post, this is not

disturbance resistance in the PCT sense. My push on the ball
bearing is not a disturbance in the PCT sense because the
position of the ball in the bowl is not a controlled
variable as can be easily determined by the fact that my push is
completely effective in changing the position of the ball.
The resistance to my push is exactly what is predicted by causal law

[From Bruce Abbott (2015.02.16.0930 EST)]

Rick Marken (2015.02.15.1510) –

RM: As I said to Bruce in an earlier post, this is not disturbance resistance in the PCT sense. My push on the ball bearing is not a disturbance in the PCT sense because the position of the ball in the bowl is not a controlled variable as can be easily determined by the fact that my push is completely effective in changing the position of the ball. The resistance to my push is exactly what is predicted by causal law

MT: Do you intend to imply that the behaviour of a control system is NOT what is predicted by causal law?

RM: No, it’s the behavior of the hypothetical controlled variable in response to disturbance that is not predicted by causal law when the variable is, indeed, controlled. For example, if we want to know whether or not the position of a mass suspended by a spring is under control we test this by applying a force disturbance, Fd, to the mass to see if it has the expected effect on the position of the mass – expected from causal law.

RM: The causal law that gives the predicted effect of the force on the position of the mass is Hooke’s law. The predicted effect of the disturbance is a change in the position of the mass, x. The predicted size of x in response to force disturbance Fd, is x = 1/s*Fd. So if we know the spring constant, s, and Fd, then we predict that applying this force to the mass will result in x amount of change in the position of the mass.

RM: If the position of the mass is not under control then the observed change in position of the mass, x’, when force Fd is applied will exactly equal x, per the causal laws of physics (approximated in Hooke’s law). If x’<<x then the position of the mass is likely under control; it is a controlled variable.

RM: This is the test for the controlled variable (TCV). Basically you are looking to see if the observed effect of a disturbance to a hypothetical controlled variable is less than expected from causal law. An understanding of causal law is essential to being able to do the TCV.

RM: Counting the restoring force exerted by the spring when Fd is applied as resistance to disturbance just confuses things because that language implies that this resistance is preventing the disturbance from having an effect, and of course it’s not.

BA: So your refusal to admit that the restoring force of an equilibrium system exerts resistance to disturbance comes down to this: You simply redefine “resistance to disturbanceâ€? as resistance that is greater than that exhibited by an equilibrium system. An equilibrium system does not, of course, produce resistance to disturbance that is greater than the resistance to disturbance produced by itself, ergo, by your idiosyncratic definition, there is “noâ€? resistance to disturbance.Â

BA: This reminds me of “newspeak� in George Orwell’s book “1984.� Newspeak was a revision of the language in which certain words meant different things depending on the reference. If I said that the enemy was enslaving the people, that meant that people were being enslaved. But if I said that the government was enslaving the people, then “enslaving� meant “freeing.� Newspeak made it impossible to say anything negative about the government.

BA:Â So, welcome to Markenspeak, where disturbances are not disturbances unless they are disturbances to a control system, and resistance to disturbance is not resistance to disturbance unless it is resistance by a control system!

Bruce

[Martin Taylor 2015.02.16.10.21]

Well, if you are going to assert quite baldly that control systems

are not part of normal science, the study of causal systems, I guess
there’s no point continuing the discussion.
In your other message [From Rick Marken (2015.02.15.1630)], you make
much the same point, asserting and reasserting that understanding
how control systems work and interact has nothing to do with
understanding PCT. The fact that the very tests of PCT that are
published as having, say, 98% agreement with what people do, is
quite irrelevant because they are studies of models and simulations.
So again, PCT is to be treated as magic, quite apart from normal
science. Discussion blocked.
Martin

···

[From Rick Marken (2015.02.15.1510)]

                        RM: As I said

to Bruce in an earlier post, this is not
disturbance resistance in the PCT sense. My
push on the ball bearing is not a
disturbance in the PCT sense because the
position of the ball in the bowl is not a
controlled variable as can be easily
determined by the fact that my push is
completely effective in changing the
position of the ball. The resistance to my
push is exactly what is predicted by causal
law

            MT: Do you intend to imply that the behaviour of

a control system is NOT what is predicted by causal law?

          RM: No, it's the

behavior of the hypothetical controlled variable in
response to disturbance that is not predicted by causal
law when the variable is, indeed, controlled.

[From Rick Marken (2015.02.16.0900)]

···

Bruce Abbott (2015.02.16.0930 EST)

 Â

RM: Counting the restoring force exerted by the spring when Fd is applied as resistance to disturbance just confuses things because that language implies that this resistance is preventing the disturbance from having an effect, and of course it’s not.Â

Â

 BA: So your refusal to admit that the restoring force of an equilibrium system exerts resistance to disturbance comes down to this: You simply redefine “resistance to disturbanceâ€? as resistance that is greater than that exhibited by an equilibrium system.  An equilibrium system does not, of course, produce resistance to disturbance that is greater than the resistance to disturbance produced by itself, ergo, by your idiosyncratic definition, there is “noâ€? resistance to disturbance.Â

RM: I’m just saying it’s confusing to call the restoring force “disturbance resistance” because it implies that equilibrium systems are like control systems, the only difference being that they don’t resist disturbances as strongly as do actual control systems. And it also seems to make the TCV impossible. Since one approach to the TCV is to look for evidence of resistance to disturbance to a hypothetical controlled variable (a la the “Coin Game” described in B:CP) then, by your definition of disturbance resistance, the position of the mass on a spring would qualify as a controlled variable since, when a force disturbance is applied to the mass there is resistance to disturbance (the restoring force) even though the mass is displaced by the expected amount.Â

RM: This is what I mean when I say that calling the restoring force “resistance to disturbance” is confusing. Doing so seems to make it impossible to distinguish between purposeful (control) and non-purposeful (causal) systems. If we call the restoring force of a spring “disturbance resistance” we would also have to call the restoring force that a book exerts against your hand when it is lifted from the table (your lift force being the disturbance to the book’s position) “disturbance resistance”. And as in the case of the mass on a spring, when the disturbance is removed the book returns to its resting state  (the “reference” state) on the table. So the position of the book, like the position of the mass on a spring, qualified as a controlled variable, by your definition of disturbance resistance. So books are control systems too.

RM: I think what equilibrium theorists are doing is using complex physics to justify making the error that William James described over 100 years ago; the error of viewing iron filings’ pursuit of a magnet as equivalent to Romeo’s pursuit of Juliet. If you block (disturb) their path and then  remove the block both continue to the goal; both were resisting the disturbance. So by your definition of disturbance resistance, iron filings and humans are the same kind of system, the difference being only one of amount (of disturbance resistance) not kind.Â

Â

BA:Â So, welcome to Markenspeak, where disturbances are not disturbances unless they are disturbances to a control system, and resistance to disturbance is not resistance to disturbance unless it is resistance by a control system!

RM: I am honored, just as I think Obama probably feels honored by Republican attempts to demonize the ACA by calling it Obamacare.Â

Best

Rick


Richard S. Marken, Ph.D.
Author of  Doing Research on Purpose
Now available from Amazon or Barnes & Noble

[From Rick Marken (2015.02.16.1230)]

Martin Taylor (2015.02.16.10.21)--

MT: Do you intend to imply that the behaviour of a control system is NOT what is predicted by causal law?

RM: No, it's the behavior of the hypothetical controlled variable in response to disturbance that is not predicted by causal law when the variable is, indeed, controlled.

MT: Well, if you are going to assert quite baldly that control systems are not part of normal science, the study of causal systems, I guess there's no point continuing the discussion... So again, PCT is to be treated as magic, quite apart from normal science. Discussion blocked.

RM: When I say causal law I assume that by now you would understand that I mean lineal or open-loop causal law rather than closed-loop causal law. I don't know if you misinterpret me because I am not clear or because you are trying to avoid dealing with the substantive parts of my posts. So let's try again. Here's the substantive part of my post to which you ignored:

RM: The causal law that gives the predicted effect of the force on the position of the mass is Hooke's law. The predicted effect of the disturbance is a change in the position of the mass, x. The predicted size of x in response to force disturbance Fd, is x = 1/s*Fd. So if we know the spring constant, s, and Fd, then we predict that applying this force to the mass will result in x amount of change in the position of the mass.
RM: If the position of the mass is not under control then the observed change in position of the mass, x', when force Fd is applied will exactly equal x, per the causal laws of physics (approximated in Hooke's law). If x'<<x then the position of the mass is likely under control; it is a controlled variable.

RM: This is the test for the controlled variable (TCV). Basically you are looking to see if the observed effect of a disturbance to a hypothetical controlled variable is less than expected from causal law. An understanding of causal law is essential to being able to do the TCV.

RM: Since you apparently took this to be my rejection of "normal science" (presumably because I am comparing the actual force-produced displacement of a mass to that expected based on "causal law") could you please explain the correct, scientific way to determine whether a mass on a spring is under control.
Best
Rick

···

--
Richard S. Marken, Ph.D.
Author of <http://www.amazon.com/Doing-Research-Purpose-Experimental-Psychology/dp/0944337554/ref=sr_1_1?ie=UTF8&qid=1407342866&sr=8-1&keywords=doing+research+on+purpose>Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

[Martin Taylor 2015.02.16.15.35]

I wish you would say "open loop" instead of "causal" when you mean

“open loop” and not “causal”.
That’s the kind of thing I mean. Furthermore, the last sentence is
wrong even under your new definition of “causal law”. You are saying
that you couldn’t do the TCV on the spring mass unless you could be
sure that it was a standard spring that conformed exactly to Hooke’s
Law. If it was a fancy spring that got stiffer as it was pulled
more, you are saying that you couldn’t do the TCV. With your
criterion, how would you do the TCV to determine whether someone was
controlling the hardness of a frying egg?
Usually when you do a test in science you are comparing one
hypothesis against another. In this case, what counter-hypothesis
has anyone else proposed? Did someone I missed suggest that maybe a
perception of a mass hanging on a spring from a fixed support is
being controlled? Who? I think nobody. Where would the controlled
perception be, anyway? So why would you want to do the TCV at all?
Martin

···

[From Rick Marken (2015.02.16.1230)]

            Martin Taylor

(2015.02.16.10.21)–

                          MT: Do you intend to imply that the

behaviour of a control system is NOT what
is predicted by causal law?

                        RM: No,

it’s the behavior of the hypothetical
controlled variable in response to
disturbance that is not predicted by causal
law when the variable is, indeed,
controlled.

            MT: Well, if you are going to assert quite baldly

that control systems are not part of normal science, the
study of causal systems, I guess there’s no point
continuing the discussion… So again, PCT is to be
treated as magic, quite apart from normal science.
Discussion blocked.

          RM: When I say causal law I assume that by now you

would understand that I mean lineal or open-loop causal
law rather than closed-loop causal law.

          I don't know if you misinterpret me because I am not

clear or because you are trying to avoid dealing with the
substantive parts of my posts. So let’s try again. Here’s
the substantive part of my post to which you ignored:

            RM: The causal law that gives the

predicted effect of the force on the position of the
mass is Hooke’s law. The predicted effect of the
disturbance is a change in the position of the mass, x.
The predicted size of x in response to force disturbance
Fd, is x = 1/s*Fd. So if we know the spring constant, s,
and Fd, then we predict that applying this force to the
mass will result in x amount of change in the position
of the mass.

            RM: If the position of the mass is not under control

then the observed change in position of the mass, x’,
when force Fd is applied will exactly equal x, per the
causal laws of physics (approximated in Hooke’s law). If
x’<<x then the position of the mass is likely
under control; it is a controlled variable.

            RM: This is the test for the

controlled variable (TCV). Basically you are looking to
see if the observed effect of a disturbance to a
hypothetical controlled variable is less than expected
from causal law. An understanding of causal law is
essential to being able to do the TCV.

            RM: Since you apparently took this to be my rejection

of “normal science” (presumably because I am comparing
the actual force-produced displacement of a mass to that
expected based on “causal law”) could you please explain
the correct, scientific way to determine whether a mass
on a spring is under control.

[From Rick Marken (2015.02.16.1600)]

···

Martin Taylor (2015.02.16.15.35)–

MT: Usually when you do a test in science you are comparing one

hypothesis against another. In this case, what counter-hypothesis
has anyone else proposed? Did someone I missed suggest that maybe a
perception of a mass hanging on a spring from a fixed support is
being controlled? Who? I think nobody. Where would the controlled
perception be, anyway? So why would you want to do the TCV at all?

RM: No wonder you don’t want to do the TCV; you don’t think it can be done.

            RM: Since you apparently took this to be my rejection

of “normal science” (presumably because I am comparing
the actual force-produced displacement of a mass to that
expected based on “causal law”) could you please explain
the correct, scientific way to determine whether a mass
on a spring is under control.

RM: Well, thanks for playing you guys, I’m done. You clearly are not interested in doing the kind of PCT research i’m interested in. Never were; never will be. But have a nice time doing your stuff.

Best

Rick

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

[Martin Taylor 2015.02.19,11 40]

I asked why you would want do the TCV to demonstrate something that

everyone has always agreed with – that the position of the mass
hanging from a spring hung from a fixed point is not being
controlled. How you turn this question into a claim that I think the
TCV cannot be done, even in that trivial case, is quite beyond me,
especially after messages in which I have proclaimed its value among
the panoply of tools available for PCT research.
Wouldn’t it be boring if all carpenters used only hammers and nails,
breaking boards by hammering, carving timbers by beating chunks out
of them, smoothing surfaces by beating down protrusions? Carpenters
use saws, chisels, screwdrivers, planes, and drills. Why should
scientists such as PCT researchers be restricted to the use of one
specific tool, and be required to use that tool and only that tool
for all research questions of any kind?
Personally, when I’m woodworking and when I am doing research, I
like to use the tool best suited for the job. And sometimes, that is
indeed the TCV.
Martin

···

[From Rick Marken (2015.02.16.1600)]

            Martin Taylor

(2015.02.16.15.35)–

                          RM: Since you apparently took this to

be my rejection of “normal science”
(presumably because I am comparing the
actual force-produced displacement of a
mass to that expected based on “causal
law”) could you please explain the
correct, scientific way to determine
whether a mass on a spring is under
control.

            MT: Usually when you do a test in science you are

comparing one hypothesis against another. In this case,
what counter-hypothesis has anyone else proposed? Did
someone I missed suggest that maybe a perception of a
mass hanging on a spring from a fixed support is being
controlled? Who? I think nobody. Where would the
controlled perception be, anyway? So why would you want
to do the TCV at all?

      RM: No wonder you don't want to do the TCV; you don't think it

can be done.

      RM: Well, thanks for playing you guys,

I’m done. You clearly are not interested in doing the kind of
PCT research i’m interested in. Never were; never will be. But
have a nice time doing your stuff.

[From Rick Marken (2015.02.19.1550)]

···

Martin Taylor (2015.02.19,11 40)–

MT: ... How you turn this question into a claim that I think the

TCV cannot be done, even in that trivial case, is quite beyond me…

RM: It’s because you simply dodged my question about how you would test to determine whether or not the position of a mass on a spring is under control by saying that it was unnecessary to do such a test in this case because you already know it is not under control. But my question clearly assumes that you don’t already know that the position of the mass is not controlled. This would be the case, for example, if you came across a mass on a spring in a control engineering lab where it is possible that the mass-spring system is connected to a control system (in a way that you can’t see) that is controlling the position of the mass. Your task is to figure out whether or not the position of the mass is under control. And the only means you have to test this is by applying disturbances to the mass.

RM: My approach was to apply known forces to the mass (by attaching known weights to the mass, say, assuming the mass is suspended vertically from the spring) and see if it had the predicted effect. The predicted effect is given by Hooke’s law, which says that the result of applying a disturbance force, Fa, to the mass is to move it by an amount x = 1/k*Fa, where k is the spring constant. If you know k then you need apply only one disturbance force to see if you get the predicted displacement, x. If you get the expected displacement then the position of the mass, x,is not under control. If you don’t get the expected displacement then it is under control.

RM: But you don’t really even need to know the spring constant,k, to do this test. You just apply several different forces to the mass and the prediction is that the displacement of the mass, x, will be linearly related to these forces, Fa, if the position of the mass, x, is not under control. If the position of the mass is under control, x will be approximately the same regardless of the value of Fa. Of course, if the spring constant is actually very large the disturbance forces (weights) will have little effect on x even if the position of the mass is not under control. So if you don’t know k and you get very little effect of different values of Fa on x then, before you conclude that x is under control you do a step in the TCV that is rarely mentioned because it’s usually obvious with a living control system: you look to see whether the lack of effect is due to the action (output) of a control system.

RM: This is the approach to the test that you deemed unscientific. So I would like to know the scientific way of determining whether or not a mass on a spring is under control.

Best

Rick

            MT: Usually when you do a test in science you are

comparing one hypothesis against another…So why would you want
to do the TCV at all?

      RM: No wonder you don't want to do the TCV; you don't think it

can be done.

                          RM: ...could you please explain the

correct, scientific way to determine
whether a mass on a spring is under
control.

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

[Martin Taylor 2015.02.19,23,02]

When did I say that it was unscientific? What I remember saying is

that scientific tests are ordiarily designed to compare one
hyothesis agains at least one competing possibility. At the time,
nobody had made any suggestion that the position of the mass on the
spring was under control. I also, on another occasion, asked Maybe that’s why you say that I claimed the method you then
described, of seeing whether the mass on a spring moves according to
your presumed law of expected effect, was an unscientific way of
determining whether it was controlled. You never answered my question, anyway. But that is normal, and to
be expected.
As for your NEW question above, the TCV has several requirements, of
which you mention one right at the very end of you penultimate
paragraph:
I agree that this is too seldom mentioned. The other stages I think
of when I do a TCV are (1) Make sure that the system has a way to
sense the putative controlled variable (in this case the position of
the mass), and (2) check that the correlation between disturbance
and the environmental equivalent of the proposed controlled
perception is low. In this example case, check that the position of
the mass is poorly related to the applied force. You didn’t mention either of these required components of the TCV,
and you added something that is not usually considered a critical
part of it, a check whether the magnitude of the effect of the
disturbance is exactly what it would be in the absence of control.
That isn’t usually a part of the TCV because it assumes that you
have some independent way to determine what the magnitude of the
effect would be, such as an assumption you make in the case of the
spring, that you know exactly the Hooke’s Law constant of the
spring. If you do have such an estimate, then of course it’s a good
short cut, but usually you don’t.
So I ask again, assuming your criterion (that an understanding of
causal law is essential to being able to do the TCV), how would you
do the TCV to determine whether someone was controlling for the
hardness of a frying egg? (Or, if that’s too easy, whether someone
was controlling for the Government to be of a particular political
party?)
Martin

···

[From Rick Marken (2015.02.19.1550)]

            Martin Taylor

(2015.02.19,11 40)–

                                        RM: ...could you please

explain the correct,
scientific way to determine
whether a mass on a spring
is under control.

                          MT: Usually when you do a test in

science you are comparing one hypothesis
against another…So why would you want to
do the TCV at all?

                    RM: No wonder you don't want to do the TCV; you

don’t think it can be done.

            MT: ... How you turn this question into a claim

that I think the TCV cannot be done, even in that
trivial case, is quite beyond me…

          RM: It's because you simply dodged my question about

how you would test to determine whether or not the
position of a mass on a spring is under control by saying
that it was unnecessary to do such a test in this case
because you already know it is not under control. But my
question clearly assumes that you don’t already know that
the position of the mass is not controlled. This would be
the case, for example, if you came across a mass on a
spring in a control engineering lab where it is possible
that the mass-spring system is connected to a control
system (in a way that you can’t see) that is controlling
the position of the mass. Your task is to figure out
whether or not the position of the mass is under control.
And the only means you have to test this is by applying
disturbances to the mass.

          RM: My approach was to apply known forces to the mass

(by attaching known weights to the mass, say, assuming
the mass is suspended vertically from the spring) and see
if it had the predicted effect. The predicted effect is
given by Hooke’s law, which says that the result of
applying a disturbance force, Fa, to the mass is to move
it by an amount x = 1/k*Fa, where k is the spring
constant. If you know k then you need apply only one
disturbance force to see if you get the predicted
displacement, x. If you get the expected displacement then
the position of the mass, x,is not under control.
If you don’t get the expected displacement then it is
under control.

          RM: But you don't really even need to know the spring

constant,k, to do this test. You just apply several
different forces to the mass and the prediction is that
the displacement of the mass, x, will be linearly related
to these forces, Fa, if the position of the mass, x, is not
under control. If the position of the mass is under
control, x will be approximately the same regardless of
the value of Fa. Of course, if the spring constant is
actually very large the disturbance forces (weights) will
have little effect on x even if the position of the mass
is not under control. So if you don’t know k and you get
very little effect of different values of Fa on x then,
before you conclude that x is under control you do a step
in the TCV that is rarely mentioned because it’s usually
obvious with a living control system: you look to see
whether the lack of effect is due to the action (output)
of a control system.

          RM: This is the approach to the test that you deemed

unscientific. So I would like to know the scientific way
of determining whether or not a mass on a spring is under
control.

    RM: This is the test for the controlled

variable (TCV). Basically you are looking to see if the observed
effect of a disturbance to a hypothetical controlled variable is
less than expected from causal law. An understanding of causal
law is essential to being able to do the TCV.

  That's the kind of thing I mean. Furthermore, the last sentence is

wrong even under your new definition of “causal law”. You are
saying that you couldn’t do the TCV on the spring mass unless you
could be sure that it was a standard spring that conformed exactly
to Hooke’s Law. If it was a fancy spring that got stiffer as it
was pulled more, you are saying that you couldn’t do the TCV. With
your criterion, how would you do the TCV to determine whether
someone was controlling the hardness of a frying egg?

  before you conclude that x is under control

you do a step in the TCV that is rarely mentioned because it’s
usually obvious with a living control system: you look to see
whether the lack of effect is due to the action (output) of a
control system

[From Rick Marken (2015.02.19.1030)]

···

Martin Taylor (2015.02.19,23,02)–

RM: You are still evading the question of how you would test to determine whether the position of a mass on a spring is a controlled. I can only assume it’s because you don’t believe that you can test to determine whether the position of the mass is controlled. Which is what I concluded in the first place; you don’t want to do research using the TCV because you don’t believe it can be done.

Best

Rick

[From Rick Marken (2015.02.19.1550)]

    RM: This is the test for the controlled

variable (TCV). Basically you are looking to see if the observed
effect of a disturbance to a hypothetical controlled variable is
less than expected from causal law. An understanding of causal
law is essential to being able to do the TCV.

  That's the kind of thing I mean. Furthermore, the last sentence is

wrong even under your new definition of “causal law”. You are
saying that you couldn’t do the TCV on the spring mass unless you
could be sure that it was a standard spring that conformed exactly
to Hooke’s Law. If it was a fancy spring that got stiffer as it
was pulled more, you are saying that you couldn’t do the TCV. With
your criterion, how would you do the TCV to determine whether
someone was controlling the hardness of a frying egg?
before you conclude that x is under control
you do a step in the TCV that is rarely mentioned because it’s
usually obvious with a living control system: you look to see
whether the lack of effect is due to the action (output) of a
control system
When did I say that it was unscientific? What I remember saying is
that scientific tests are ordiarily designed to compare one
hyothesis agains at least one competing possibility. At the time,
nobody had made any suggestion that the position of the mass on the
spring was under control.

I also, on another occasion, asked




Maybe that's why you say that I claimed the method you then

described, of seeing whether the mass on a spring moves according to
your presumed law of expected effect, was an unscientific way of
determining whether it was controlled.

You never answered my question, anyway. But that is normal, and to

be expected.

As for your NEW question above, the TCV has several requirements, of

which you mention one right at the very end of you penultimate
paragraph:

I agree that this is too seldom mentioned. The other stages I think

of when I do a TCV are (1) Make sure that the system has a way to
sense the putative controlled variable (in this case the position of
the mass), and (2) check that the correlation between disturbance
and the environmental equivalent of the proposed controlled
perception is low. In this example case, check that the position of
the mass is poorly related to the applied force.

You didn't mention either of these required components of the TCV,

and you added something that is not usually considered a critical
part of it, a check whether the magnitude of the effect of the
disturbance is exactly what it would be in the absence of control.
That isn’t usually a part of the TCV because it assumes that you
have some independent way to determine what the magnitude of the
effect would be, such as an assumption you make in the case of the
spring, that you know exactly the Hooke’s Law constant of the
spring. If you do have such an estimate, then of course it’s a good
short cut, but usually you don’t.

So I ask again, assuming your criterion (that an understanding of

causal law is essential to being able to do the TCV), how would you
do the TCV to determine whether someone was controlling for the
hardness of a frying egg? (Or, if that’s too easy, whether someone
was controlling for the Government to be of a particular political
party?)

Martin

            Martin Taylor

(2015.02.19,11 40)–

            MT: ... How you turn this question into a claim

that I think the TCV cannot be done, even in that
trivial case, is quite beyond me…

          RM: It's because you simply dodged my question about

how you would test to determine whether or not the
position of a mass on a spring is under control by saying
that it was unnecessary to do such a test in this case
because you already know it is not under control. But my
question clearly assumes that you don’t already know that
the position of the mass is not controlled. This would be
the case, for example, if you came across a mass on a
spring in a control engineering lab where it is possible
that the mass-spring system is connected to a control
system (in a way that you can’t see) that is controlling
the position of the mass. Your task is to figure out
whether or not the position of the mass is under control.
And the only means you have to test this is by applying
disturbances to the mass.

          RM: My approach was to apply known forces to the mass

(by attaching known weights to the mass, say, assuming
the mass is suspended vertically from the spring) and see
if it had the predicted effect. The predicted effect is
given by Hooke’s law, which says that the result of
applying a disturbance force, Fa, to the mass is to move
it by an amount x = 1/k*Fa, where k is the spring
constant. If you know k then you need apply only one
disturbance force to see if you get the predicted
displacement, x. If you get the expected displacement then
the position of the mass, x,is not under control.
If you don’t get the expected displacement then it is
under control.

          RM: But you don't really even need to know the spring

constant,k, to do this test. You just apply several
different forces to the mass and the prediction is that
the displacement of the mass, x, will be linearly related
to these forces, Fa, if the position of the mass, x, is not
under control. If the position of the mass is under
control, x will be approximately the same regardless of
the value of Fa. Of course, if the spring constant is
actually very large the disturbance forces (weights) will
have little effect on x even if the position of the mass
is not under control. So if you don’t know k and you get
very little effect of different values of Fa on x then,
before you conclude that x is under control you do a step
in the TCV that is rarely mentioned because it’s usually
obvious with a living control system: you look to see
whether the lack of effect is due to the action (output)
of a control system.

          RM: This is the approach to the test that you deemed

unscientific. So I would like to know the scientific way
of determining whether or not a mass on a spring is under
control.

                          MT: Usually when you do a test in

science you are comparing one hypothesis
against another…So why would you want to
do the TCV at all?

                    RM: No wonder you don't want to do the TCV; you

don’t think it can be done.

                                        RM: ...could you please

explain the correct,
scientific way to determine
whether a mass on a spring
is under control.

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

PY:

I am under the impression that the TCV is an analysis of the shared environment of two control systems, one of which is a scientist.

RM:

Basically you are looking to see if the observed effect of a disturbance to a hypothetical controlled variable is less than expected from causal law

PY:

If we push on a spring with F, the effect is to displace the spring by F/k.

MT:

[To do the TCV] in this example case, check that the position of the mass is poorly related to the applied force.

PY:

The position of the mass on the spring is in the shared environment of two hypothetical control systems, one of which is a scientist.
The scientist applies F and observes a displacement equal or slightly in error of that predicted by Hooke’s Law. After repeated application of the same F leads to the same displacement, the validity of Hooke’s Law is no longer the issue.

MT:

before you conclude that x is under control, you…look to see whether the lack of effect is due to the action (output) of a[nother] control system.

PY:

The scientist who is doing the TCV on the position of the mass on the spring applies a hypothetical disturbance to this position by controlling it. In the absence of this disturbance, the scientist knows that the spring would assume its equilibrium position, and therefore must accept that the position of the mass is under control by at least one control system. To do the TCV proper, the scientist must check whether another hypothetical control system (say, a visible or invisible human, located in proximity) is also controlling the position of the mass and is responding to our control (the disturbance). The TCV is, therefore, checking to see whether a reference condition in one control system is the cause of the disturbance to another control system’s perception.

MT:

[If] the correlation between disturbance and the environmental equivalent of the proposed controlled perception is low, [the TCV fails].

PY:

In any case where two control systems are not pitted against each other (one of which is a scientist attempting to disturb a hypothetical controlled variable), the TCV does not apply. In this case, I believe there is only one “person” applying a force to the spring, so the TCV does not apply.

kind regards,

Philip

···

[philip 9:45.2/20/2015]

[Martin Taylor 2015.02.20.15.20]

If you have an independent measure of k, you can do this. If you

don’t, you can’t. How do you do it when the question is, as in one
of Rick’s example demos, whether when someone is asked to control
the area of a rectangle they control x+y, 0.9x+1.1y, x^2+y^2, xy,
(x
y)^0.75, log(x*y), the maximum distance from an internal point to
the nearest edge, or something other function? Or, as I asked Rick,
if you want to know whether they are controlling the harness of a
frying egg? What you can do is compare the relative likelihoods that
perceptions formed from the different proposed perceptual input
functions are the one that is actually being controlled. The way you
do it is, as Rick points out in
:
"
."
Incidentally, Rick, although it’s over six years since you sy you
changed this from the area of a square, you still say “square” at
the end of this quote.
OK, except for “knows”, since it assumes that the scientist has
somehow determined that the thing being tested is in fact a spring,
and that the top of the spring is on a fixed support. That
assumption rather voids the point of the test, which is to determine
whether the position of the mass is being controlled to oppose the
disturbance applied by the scientist.
But how does this follow?
For “fails” please substitute “succeeds”. The signature of control
is that the CEV variation for good control is uncorrelated with the
disturbance or nearly so. The output of a good control system is highly negatively correlated
with the disturbance. Using the ordinary geometric representation of
correlation and taking o and d as the two influences on the CEV, the
vectors o and -d are at a small angle to each other and of similar
magnitude. Their difference vector, the variation of the supposed
CEV, whose magnitude relative to the magnitude of the disturbance
vector is the measure of control quality, is at a probabilistically
distributed angle distributed around orthogonality. If you have the
habit of seeing correlations geometrically, this should be
immediately apparent to you. If you don’t, you can work out the same
thing algebraically. There’s a slight chance that the CEV variation
is correlated with the disturbance, This happens when the outut is
perfectly correlated with the disturbance but is of different
magnitude, a situation unlikely to persist.
That’s actually the case in which it is almost always applied.
The question Rick raised in his last message was whether there is
indeed only one person, the scientist, controlling the position of
the mass. The answeris “No” if the mass position is highly
correlated with the disturbance applied by the scientist. There is no control in this (or any other) equilibrium system,
though control systems can incorporate equilibrium systems in their
environmental feedback path.
Martin

···

[philip 9:45.2/20/2015]

PY:

I am under the impression that the TCV is a n
analysis of the shared environment of two control systems, one
of which is a scientist.

RM:

              Basically

you are looking to see if the observed effect of a
disturbance to a hypothetical controlled variable is
less than expected from causal law

PY:

  If we push on a spring with F, the effect is to displace the

spring by F/k.

MT:

        [To

do the TCV] in this example case, check that the position of
the mass is poorly related to the applied force.

PY:

  The position of the mass on the spring is in the shared

environment of two hypothetical control systems, one of which is a
scientist.
The scientist applies F and observes a displacement equal or
slightly in error of that predicted by Hooke’s Law. After
repeated application of the same F leads to the same
displacement, the validity of Hooke’s Law is no longer the
issue.

http://www.mindreadings.com/ControlDemo/Mindread.html

  •  The computer is able to determine your purpose (to read your
    

mind) by doing a version of the control theory based “test for the
controlled variable”. The movements of all three characters are
potential* controlled variables* .
The computer determines the controlled character (the one moved on
purpose) by continuously calculating the correlation between the
movements of each character and the* disturbance* to each character. A different time varying disturbance is applied
to each character. In order to keep the movements of the
controlled character under control, you have to oppose the effects
of the disturbance to that character. This means that the
correlation between the character’s movement and the disturbance
to those movements will be* lowestfor the controlled square

MT:

          before you

conclude that x is under control, you…look to see
whether the lack of effect is due to the action
(output) of a[nother] control system.

PY:

    The scientist who is doing the TCV on the

position of the mass on the spring applies a
hypothetical disturbance to this position by controlling it. In
the absence of this disturbance, the scientist knows that the
spring would assume its equilibrium position,

    and therefore must accept that the position of the mass is

under control by at least one control system.

    To do the TCV proper, the scientist must check whether

another hypothetical control system (say, a visible or invisible
human, located in proximity) is also controlling the position of
the mass and is responding to our control (the disturbance). The
TCV is, therefore, checking to see whether a reference condition
in one control system is the cause of the disturbance to another
control system’s perception.

MT:

[If] the
correlation between disturbance and the environmental
equivalent of the proposed controlled perception is low, [the
TCV fails].

PY:

    In any case where two control systems are not pitted against

each other (one of which is a scientist attempting to disturb a
hypothetical controlled variable), the TCV does not apply.

    In this case, I believe there is only one "person" applying

a force to the spring, so the TCV does not apply.

[Martin Taylor 2015.02.20.15.55]

[From Rick Marken (2015.02.19.1030)]

Am I? Let’s see…

[Martin Taylor 2015.02.19,23,02]
That’s the message you claim to be responding to. Here is what I
believe to be a statement of how the TCV should be done, you having
already pointed out that the TCV really also include determining
whether there is an output that could influence the presumed CEV.
The other stages I think of when I do a TCV
are (1) Make sure that the system has a way to sense the putative
controlled variable (in this case the position of the mass), and
(2) check that the correlation between disturbance and the
environmental equivalent of the proposed controlled perception is
low. In this example case, check that the position of the mass is
poorly related to the applied force.

Note the last sentence. That's the actual Test.

I guess you read my messages as carefully as you read Bruce's, and

use a similar translation method to render them into Markenese for
interpretation.

Martin
···
            Martin Taylor

(2015.02.19,23,02)–

          RM: You are still evading the question of how you would

test to determine whether the position of a mass on a
spring is a controlled. I can only assume it’s because
you don’t believe that you can test to determine whether
the position of the mass is controlled. Which is what I
concluded in the first place; you don’t want to do
research using the TCV because you don’t believe it can be
done.

[From Rick Marken (2015.02.20.1915)]

···

Martin Taylor ( 2015.02.20.15.55)–

  MT: The other stages I think of when I do a TCV

are (1) Make sure that the system has a way to sense the putative
controlled variable (in this case the position of the mass), and
(2) check that the correlation between disturbance and the
environmental equivalent of the proposed controlled perception is
low. In this example case, check that the position of the mass is
poorly related to the applied force.
MT: Am I? Let’s see…
MT: Note the last sentence. That’s the actual Test.

RM: Close. You have to say what “poorly related” means. You should also say what “well related” means. Using your terminology, the test to determine whether the position of a mass on a spring is under control is to look to see if this variable is “poorly related” to the disturbance (the force applied to the mass). If the position of the mass is “poorly related” to the disturbance then that is evidence that the position of the mass is under control. If, however, the position of the mass is “well related” to the disturbance then the position of the mass is not under control.

RM: Fill in the correct substitutions for “poorly related” and “well related” and I will be convinced that you know how to test to determine whether a mass on a spring is controlled or not.

Best

Rick

          RM: You are still evading the question of how you would

test to determine whether the position of a mass on a
spring is a controlled. I can only assume it’s because
you don’t believe that you can test to determine whether
the position of the mass is controlled. Which is what I
concluded in the first place; you don’t want to do
research using the TCV because you don’t believe it can be
done.

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble