FW: Control Theory Entry in APA Dictionary of Psychology

[_Boris_Hartman]

Rick,

I see that you are trying hard to change your standpoint. It’s worthy of praise. But I think it’s not enough. You are still too much in »objective« world. There is so little »Control of perception«. Some things are wrong again. If you want to know why you’ll have to check our previous conversations and with Martin. Sorry I don’t have time to repeat things so many times. And I advice you to read some of Bill’s literature, because you obviously overlooked some important informations.

I’ll wait what you will write to APA J.

Best,

Boris

···

From: Richard Marken (rsmarken@gmail.com via csgnet Mailing List) [mailto:csgnet@lists.illinois.edu]
Sent: Sunday, March 22, 2015 8:43 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control Theory Entry in APA Dictionary of Psychology

[From Rick Marken (2015.03.22.1245)]

On Sun, Mar 22, 2015 at 10:45 AM, “Boris Hartman” csgnet@lists.illinois.edu wrote:

HB Behavior is NOT CONTROL. It is means of control.

RM: OK, so you want to restrict the word “behavior” to denote only variations in the actions that keep a controlled variable at its reference.

HB : You are wrong. I don’t want to restrict anything. Bill did.

RM: I don’t think he did. He shows that the things we call behaviors involve varying actions to produce stable (controlled) results; so a behavior such as “walking” involves producing a stable result – the walk – by varying actions appropriately – the leg movements that result in the walk.

HB: Which »controlled variable« is kept at it’s reference ?

RM: It depends on what behavior you are talking about. There are many controlled variables involved in walking, for example, but perhaps the highest level controlled variable is the person’s change of position from one place to another. This is accomplished by movements of the legs, which are the actions that result in the change in position. These leg movements are themselves controlled variables, the actions that result in control being the forces exerted by the muscles on the bones of the legs. There is clearly a hierarchy of controlled variables (and the actions that control them) that make up what we call “behavior”. This is why Bill developed the hierarchical control model of behavior.

RM: Then what do you call the variations in the controlled variable itself?

HB : First you must tell me which »controlled variable« you have in mind.

RM I already explained it above so you’re off the hook.

HB: Your confusion with »controlled variables« is well known. What does it mean »controlled varibel itself« ?

RM: A controlled variable is a perceptual representation of aspects of a control system’s sensory input. It defines the aspect of the environment that the control system controls.

HB: Something outside ?

RM: Not necessarily. It could be a function of what is outside. For example, a control system that controls humidity controls a perception of something that doesn’t exist outside the system. Humidity is a perceptual function of the sensed temperature and water vapor content of the air. So while humidity is not a variable that exists “outside” the humidity control system, it is a variable that is a function of variables that do exist outside the control system. So it is possible for an outside observer to perceive what the control system is perceiving (using the appropriate measuring devices) and determine that the system is controlling a perception of humidity.

HB: I’m also interested what you meant by »variations in controlled variable ITSELF« ? You know what is »controlled variable« itself ?

RM: Control systems – especially living control systems – often maintain controlled variables in variable reference states, usually as the means of controlling higher order perceptions. This is what is going on in walking, where the position of the legs is a controlled variable that is varied in order to control a person’s balance as well as control for moving the person from one place to another.

HB: But whatever you are writing, must be your definiton of »behavior« as it turned out in your past discussions with me and your discussion with Martin. You could go and read it again, what you think behavior is. It suits your »control of behavior«.

RM: I defined “behavior” as I am defining it now – as control – in my 1988 paper “The Nature of Behavior: Control as Fact and Theory”. It’s reprinted as the first paper in “Mind Readings”. I recommend it to you.

HB: So I suppose that whatever you described above is your definition of behavior as it fits quite good in self-regulation which you promoted many times, only GOALS are missing.

RM: “Goal” is an informal term (like behavior) but the concept of goal is implicit in the definition of behavior as control since control involves maintaining variables in (possibly variable) “goal” states.

HB: Where is »Control of perception« ? Which perceptions walker controls ?

RM: Excellent question and it is one that can only be answered by PCT research. Indeed, this is the fundamental question of PCT research. We can tell from observation that the behavior we call walking is a control process; there are variables, such as the person’s forward progress and balance, that are clearly under control – they are stable patterns of behavior that would not happen due to disturbances – such as unevenness of the ground – without compensating actions – such as appropriately placed legs. So the problem is to figure out what variables are being controlled from the control system’s perspective; what perceptual aspects of its environment is the system controlling when it walks. This is where the test for the controlled variable comes in; it’s also where the robotics research comes in because the people building robots have to figure out what perceptions the robot should control in order to be able to produce the behavior we call walking. I would guess that the variables controlled when walking include optical variables (such as one’s orientation relative to buildings and such), proprioceptive variables (such as the feeling of orientation of the foot) and kinesthetic variables (such as rate of change in position of the leg).

RM: Do you think that is the meaning of “behavior” Bill Powers had in mind when he wrote “Behavior: The Control of Perception”?

HB : I don’t know what Bill had in mind. Do you ?

RM: It looks like he had in mind that what we are seeing when we see “behavior” is the process of a person acting to control their own perceptions.

HB: But I can try to conclude what he thought from the Title : »Behavior : Control of Perception«. He probably meant exactly what he wrote. What do you think he meant ?

RM: Yes, I agree.

HB: If I recal from your conversation with Martin, you think that »Perception is Control of Behavior«.

RM: That’s just silly. All of my work is aimed at showing that behavior is the control of perception and that the appearance that perception controls behavior is an illusion.

HB :

….It seems that you made two prremises from Bill’s theory :

1.   Behavior is control
   

…and next Bill’s most important realization was that

2.   these systems control by controlling their inputs not their outputs. 
   

These are your words.

RM: Yes.

HB: Can you explain what all this means ? Can we really reduce Bill’s theory only to these two premises ?

RM: I think so. Of course, it takes quite a bit to explain what they mean. There’s a lot of meaning packed into the little phase “Behavior is control”. You have to know what control is and how it relates to the things we call “behaviors”. Bill did that in his 1979 BYTE article titled “Defining Behavior” (http://www.livingcontrolsystems.com/enclosures/byte_june_1979.pdf) and I did it in my 1988 “Nature of Behavior” paper, mentioned above. And, of course, there is the whole of control theory packed into that second phrase “control systems control their inputs, not their outputs”.

HB: Could you explain how can we »understand the nature of living systems« from these two premises ?

RM: By reading everything Bill wrote about PCT and by doing Bill’s demos. I would also suggest that you read my three books on control theory as well (Mind Readings, More Mind Readings and Doing Research on Purpose) and do all my demos (http://www.mindreadings.com/demos.htm).

HB: And still I think that whatever you are up to when you intend to promote yourself (I suppose that you will write your name under what you will send to APA) instead of promoting PCT, it’s good to talk about first here on CSGnet. Otherwise who knows what can happen.

RM: Why don’t you write a version and we can see what what you think it should look like.

HB: And I still think that it’s better that somebody else help with his knowledge. Like Kent for example as he already I think on 2011 CSG mitting. Maybe he can help in times when you are so busy J.

RM: I’m going to write it just with Warren Mansell. I think too many cooks might spoil the broth.

Best

Rick

–

Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.

Now available from Amazon or Barnes & Noble

[rsmarken]

[From Rick Marken (2015.03.22.1445)

···

On Sun, Mar 22, 2015 at 1:53 PM, “Boris Hartman” csgnet@lists.illinois.edu wrote:

Rick,

Â

HB: I see that you are trying hard to change your standpoint. It’s worthy of praise. But I think it’s not enough. You are still too much in »objective« world. There is so little »Control of perception«. Some things are wrong again. If you want to know why you’ll have to check our previous conversations and with Martin.

RM: I would prefer that you explain the things that you think I got wrong yourself.Â

Â

HB: Sorry I don’t have time to repeat things so many times.

RM: I don’t see what would take so much time. All I’ve ever heard you say is “we only control perception”, which is true enough. You seem to think that I am saying something different. But I’m not. We only control perception. But perceptions are functions of environmental variables. So when we control perceptions we are controlling aspects of the environment that correspond to those perceptions. You don’t seem to like that idea but I think you would, if not like it better, at least understand what it means if you would learn PCT by building models and/or robots to imitate some purposeful behavior. If your understanding of PCT comes only from words then you are really working with one hand  (and possibly two) tied behind your back.Â

HB: And I advice you to read some of Bill’s literature, because you obviously overlooked some important informations.

RM: I’ve read all of Bill’s work but, more importantly, I’ve done all of Bill demos. But I really started understanding PCT only after doing the computer modeling (which I started doing in 1979 based on the last article in the BYTE series; http://www.livingcontrolsystems.com/enclosures/byte_sep_1979.pdf). There are also some very nice examples of building simple control models in “LCS III: The fact of control”. I highly recommend that you not only read that book but also build the computer models. If you can’t build the models at least go through the logic of them; it might help you understand PCT a bit better than you apparently do now. What you will find is, that in order to build a model of, say, a tracking task, you will have to have a statement in the model that says something like this:Â

p:= t - c

RM: What this says is that the perceptual variable under control, p, is a function of two environmental variables, t, the target position, and c, the cursor position. So when the control system controls p, making p = r, it is also controlling t-c. So when we say behavior is the “control of perception” we are also saying that behavior is the control of aspects of the environment that correspond to the perception. That’s what keeps PCT from being a nonsensical exercise in solipsism.Â

 HB: I’ll wait what you will write to APA J.

RM: Oh, don’t wait Boris. This is too much fun.

BestÂ

Rick

Â

Â

Best,

Â

Boris

Â

From: Richard Marken (rsmarken@gmail.com via csgnet Mailing List) [mailto:csgnet@lists.illinois.edu]
Sent: Sunday, March 22, 2015 8:43 PM
To: csgnet@lists.illinois.edu
Subject: Re: Control Theory Entry in APA Dictionary of Psychology

Â

[From Rick Marken (2015.03.22.1245)]

Â

On Sun, Mar 22, 2015 at 10:45 AM, “Boris Hartman” csgnet@lists.illinois.edu wrote:

Â

 HB  Behavior is NOT CONTROL. It is means of control.

Â

RM: OK, so you want to restrict the word “behavior” to denote only variations in the actions that keep a controlled variable at its reference.

Â

HB : You are wrong. I don’t want to restrict anything. Bill did.

Â

RM: I don’t think he did. He shows that the things we call behaviors involve varying actions to produce stable (controlled) results; so a behavior such as “walking” involves producing a stable result – the walk – by varying actions appropriately – the leg movements that result in the walk.Â

Â

HB: Which »controlled variable« is kept at it’s reference ?

Â

RM: It depends on what behavior you are talking about. There are many controlled variables involved in walking, for example, but perhaps the highest level controlled variable is the person’s change of position from one place to another. This is accomplished by movements of the legs, which are the actions that result in the change in position. These leg movements are themselves controlled variables, the actions that result in control being the forces exerted by the muscles on the bones of the legs. There is clearly a hierarchy of controlled variables (and the actions that control them) that make up what we call “behavior”. This is why Bill developed the hierarchical control model of behavior.Â

Â

RM: Then what do you call the variations in the controlled variable itself?Â

Â

HB : First you must tell me which »controlled variable« you have in mind.

Â

RM I already explained it above so you’re off the hook.Â

Â

HB: Your confusion with »controlled variables« is well known. What does it mean »controlled varibel itself« ?

Â

RM: A controlled variable is a perceptual representation of aspects of a control system’s sensory input. It defines the aspect of the environment that the control system controls.

Â

HB: Something outside ?

Â

RM: Not necessarily. It could be a function of what is outside. For example, a control system that controls humidity controls a perception of something that doesn’t exist outside the system. Humidity is a perceptual function of the sensed temperature and water vapor content of the air. So while humidity is not a variable that exists “outside” the humidity control system, it is a variable that is a function of variables that do exist outside the control system. So it is possible for an outside observer to perceive what the control system is perceiving (using the appropriate measuring devices) and determine that the system is controlling a perception of humidity.

Â

HB: I’m also interested what you meant by »variations in controlled variable ITSELF« ? You know what is »controlled variable« itself ?

Â

RM: Control systems – especially living control systems – often maintain controlled variables in variable reference states, usually as the means of controlling higher order perceptions. This is what is going on in walking, where the position of the legs is a controlled variable that is varied in order to control a person’s balance as well as control for moving the person from one place to another.Â

Â

HB: But whatever you are writing, must be your definiton of »behavior« as it turned out in your past discussions with me and your discussion with Martin. You could go and read it again, what you think behavior is. It suits your »control of behavior«.

Â

RM: I defined “behavior” as I am defining it now – as control – in my 1988 paper “The Nature of Behavior: Control as Fact and Theory”. It’s reprinted as the first paper in “Mind Readings”. I recommend it to you.Â

Â

HB: Â So I suppose that whatever you described above is your definition of behavior as it fits quite good in self-regulation which you promoted many times, only GOALS are missing.

Â

RM: “Goal” is an informal term (like behavior) but the concept of goal is implicit in the definition of behavior as control since control involves maintaining variables in (possibly variable) “goal” states.Â

Â

HB: Where is »Control of perception« ? Which perceptions walker controls ?

Â

RM: Excellent question and it is one that can only be answered by PCT research. Indeed, this is the fundamental question of PCT research. We can tell from observation that the behavior we call walking is a control process; there are variables, such as the person’s forward progress and balance, that are clearly under control – they are stable patterns of behavior that would not happen due to disturbances – such as unevenness of the ground – without compensating actions – such as appropriately placed legs. So the problem is to figure out what variables are being controlled from the control system’s perspective; what perceptual aspects of its environment is the system controlling when it walks. This is where the test for the controlled variable comes in; it’s also where the robotics research comes in because the people building robots have to figure out what perceptions the robot should control in order to be able to produce the behavior we call walking. I would guess that the variables controlled when walking include optical variables (such as one’s orientation relative to buildings and such), proprioceptive variables (such as the feeling of orientation of the foot) and kinesthetic variables (such as rate of change in position of the leg).Â

Â

RM:Â Do you think that is the meaning of “behavior” Bill Powers had in mind when he wrote “Behavior: The Control of Perception”?Â

Â

HB : I don’t know what Bill had in mind. Do you ?

Â

RM: It looks like he had in mind that what we are seeing when we see “behavior” is the process of a person acting to control their own perceptions.Â

Â

Â

HB: But I can try to conclude what he thought from the Title : »Behavior : Control of Perception«. He probably meant exactly what he wrote. What do you think he meant ?

Â

 RM: Yes, I agree.Â

Â

HB: If I recal from your conversation with Martin, you think that »Perception is Control of Behavior«.

Â

RM: That’s just silly. All of my work is aimed at showing that behavior is the control of perception and that the appearance that perception controls behavior is an illusion.Â

Â

HB :

….It seems that you made two premises from Bill’s theory :

1.      Behavior is control

…and next Bill’s most important realization was that

2.      these systems control by controlling their inputs not their outputs.Â

These are your words.

Â

RM: Yes.Â

Â

 HB: Can you explain what all this means ? Can we really reduce Bill’s theory only to these two premises ?

Â

RM: I think so. Of course, it takes quite a bit to explain what they mean. There’s a lot of meaning packed into the little phase “Behavior is control”. You have to know what control is and how it relates to the things we call “behaviors”. Bill did that in his 1979 BYTE article titled “Defining Behavior” (http://www.livingcontrolsystems.com/enclosures/byte_june_1979.pdf) and I did it in my 1988 “Nature of Behavior” paper, mentioned above. And, of course, there is the whole of control theory packed into that second phrase “control systems control their inputs, not their outputs”.

Â

HB: Could you explain how can we »understand the nature of living systems« from these two premises ?

Â

RM: By reading everything Bill wrote about PCT and by doing Bill’s demos. I would also suggest that you read my three books on control theory as well (Mind Readings, More Mind Readings and Doing Research on Purpose) Â and do all my demos (http://www.mindreadings.com/demos.htm).Â

Â

 HB: And still I think that whatever you are up to when you intend to promote yourself (I suppose that you will write your name under what you will send to APA) instead of promoting PCT, it’s good to talk about first here on CSGnet. Otherwise who knows what can happen.

Â

RM: Why don’t you write a version and we can see what what you think it should look like.Â

Â

HB: And I still think that it’s better that somebody else help with his knowledge. Like Kent for example as he already I think on 2011 CSG mitting. Maybe he can help in times when you are so busy  J.

Â

RM: I’m going to write it just with Warren Mansell. I think too many cooks might spoil the broth.Â

Â

BestÂ

Â

Rick

Â

–

Richard S. Marken, Ph.D.
Author of  Doing Research on Purpose.Â

Now available from Amazon or Barnes & Noble

–

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

[FredNickols]

[From Fred Nickols (2015.03.23.0630 EDT)]

[From Rick Marken (2015.03.22.1445)

RM: I don’t see what would take so much time. All I’ve ever heard you say is “we only control perception”, which is true enough. You seem to think that I am saying something different. But I’m not. We only control perception. But perceptions are functions of environmental variables. So when we control perceptions we are controlling aspects of the environment that correspond to those perceptions. You don’t seem to like that idea but I think you would, if not like it better, at least understand what it means if you would learn PCT by building models and/or robots to imitate some purposeful behavior. If your understanding of PCT comes only from words then you are really working with one hand (and possibly two) tied behind your back.

[Fred Nickols]

So behavior is also the control of aspects of our environment? How does that jibe with “We only control perception.�?

Regards,

Fred Nickols

Distance Consulting LLC

www.nickols.us

[MartinT]

[Martin Taylor 2015.03.23.09.44]

Rick's statement might be less controversial if on the third line he

had added a couple of extra words, so that the sentence would read
“So when we control perceptions, we seem to be controlling aspects
of the environment that correspond to those perceptions.” This is
true simply because the perceptual function defines how
environmental changes influence the perceptual value. If perception
= P(environment), then if P( ) is monotonic, environment =
P^-1(perception). Because of the inevitable time delay between a change in the
environment and its influence on the perception, if P( ) is causal,
then P^-1( ) is non-causal, which means non-physical. The perception
does not cause anything to happen in the environment, but this
does not matter, because we are only trying to determine what
perception is being controlled. If the time-lag between
environmental change and the corresponding change in perception is
small compared to the loop transport lag, we can treat the
environmental function as being equivalent to the perception, and
therefore as being controlled, even though it is technically not
being controlled.
The only access we have to the perceptions others may be controlling
is the “Test for the Controlled Variable” (TCV). The TCV depends (as
it must) on effects in the world observable by the tester. The
tester can never have exactly the same perceptions of the world as
the subject of the Test, but in many cases, especially for low-level
perceptions, the two sets of perceptions can be assumed to be pretty
close. So, when testing a parametric range of possible candidates
for the controlled perception, the one that results in the
appearance of best control of the hypothesised environmental
function is the best bet for a definition of the perceptual
function. Of course, it is always possible that what is actually
being controlled includes some variable not incorporated into the
test, but the “unknown unknown” can never be included. We do the
best with what we can imagine.
The situation is different when the tester wants only to determine
which of a finite number of pre-defined possible perceptions is
being controlled, as in Rick’s demos. Which is the car whose
location you are controlling, of three possibilities? The answer is
quite clear, without trying to answer exactly what defines the
perception of “location” that you are controlling.
Anway, the point is that for many purposes one can treat the
environmental correlate of the perceptual function as though it were
being controlled exactly as the corresponding perception is
controlled, but when you come to detail you have to be a bit
cautious about it, especially when the candidate perceptions are at
a high level, as Robertson et al. found when they used the TCV in a
study of self-image (Robertson, R. J., Goldstein, D. M., Mermel, M.,
& Musgrave, M. (1999). Testing the self as a control system:
Theoretical and methodological issues. International journal of
human-computer studies, 50, 571-580).
Martin

···

        [From

Fred Nickols (2015.03.23.0630 EDT)]

Â

                [From Rick Marken

(2015.03.22.1445)

Â

                RM: I don't see what would take

so much time. All I’ve ever heard you say is “we
only control perception”, which is true enough. You
seem to think that I am saying something different.
But I’m not. We only control perception. But
perceptions are functions of environmental
variables. So when we control perceptions we are
controlling aspects of the environment that
correspond to those perceptions. You don’t seem to
like that idea but I think you would, if not like it
better, at least understand what it means if you
would learn PCT by building models and/or robots to
imitate some purposeful behavior. If your
understanding of PCT comes only from words then you
are really working with one hand  (and possibly two)
tied behind your back.

Â

*** [Fred
Nickols]***

*** So
behavior is also the control of aspects of our
environment? How does that jibe with “We only
control perception.�?***

Â

[rsmarken]

[From Rick Marken (2015.03.23.0930)]

···

 Fred Nickols (2015.03.23.0630 EDT)

Â

[From Rick Marken (2015.03.22.1445)

Â

RM: I don’t see what would take so much time. All I’ve ever heard you say is “we only control perception”, which is true enough. You seem to think that I am saying something different. But I’m not. We only control perception. But perceptions are functions of environmental variables. So when we control perceptions we are controlling aspects of the environment that correspond to those perceptions. You don’t seem to like that idea but I think you would, if not like it better, at least understand what it means if you would learn PCT by building models and/or robots to imitate some purposeful behavior. If your understanding of PCT comes only from words then you are really working with one hand  (and possibly two) tied behind your back.

Â

[Fred Nickols]

So behavior is also the control of aspects of our environment? How does that jibe with “We only control perception.�?

RM: I don’t know if I can make it any clearer than I did in a later part of that post:Â

RM: What you will find  [when you do the modeling] is, that in order to build a model of, say, a tracking task, you will have to have a statement in the model that says something like this:Â
Â

p:= t - c
Â

RM: What this says is that the perceptual variable under control, p, is a function of two environmental variables, t, the target position, and c, the cursor position. So when the control system controls p, making p = r, it is also controlling t-c. So when we say behavior is the “control of perception” we are also saying that behavior is the control of aspects of the environment that correspond to the perception. That’s what keeps PCT from being a nonsensical exercise in solipsism.Â
Â

RM: Does that help?Â

BestÂ

Rick

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

[rsmarken]

[From Rick Marken (2015.03.23.0945)]

···

Martin Taylor (2015.03.23.09.44)–

*** [Fred
Nickols]***

*** So
behavior is also the control of aspects of our
environment? How does that jibe with “We only
control perception.�?***

MT: Rick's statement might be less controversial if on the third line he

had added a couple of extra words, so that the sentence would read
“So when we control perceptions, we seem to be controlling aspects
of the environment that correspond to those perceptions.”

RM: This certainly does describe the phenomenology. We experience our perceptions of the environment as the environment. I, for example, experience the computer I’m typing on as an actual thing out there in the environment, not as a perception. It is a perception, of course, but I don’t experience it that way.Â

RM: But that wasn’t my point. My point was exactly as I said it: When we control perceptions we are also controlling aspects of the environment that correspond to those perceptions. Perhaps to get my point across more clearly I should have said: When we control perceptions we are also controlling aspects of the environment (per the current models of physics and chemistry) that correspond to those perceptions. It’s true that the “real” environment is known to us only as a theory – a well established theory but a theory nevertheless. But that theory – the models of physics and chemistry – is part of PCT; it’s the “environment” side of the model of a PCT control system.Â

BestÂ

Rick

Because of the inevitable time delay between a change in the

environment and its influence on the perception, if P( ) is causal,
then P^-1( ) is non-causal, which means non-physical. The perception
does not cause anything to happen in the environment, but this
does not matter, because we are only trying to determine what
perception is being controlled. If the time-lag between
environmental change and the corresponding change in perception is
small compared to the loop transport lag, we can treat the
environmental function as being equivalent to the perception, and
therefore as being controlled, even though it is technically not
being controlled.

The only access we have to the perceptions others may be controlling

is the “Test for the Controlled Variable” (TCV). The TCV depends (as
it must) on effects in the world observable by the tester. The
tester can never have exactly the same perceptions of the world as
the subject of the Test, but in many cases, especially for low-level
perceptions, the two sets of perceptions can be assumed to be pretty
close. So, when testing a parametric range of possible candidates
for the controlled perception, the one that results in the
appearance of best control of the hypothesised environmental
function is the best bet for a definition of the perceptual
function. Of course, it is always possible that what is actually
being controlled includes some variable not incorporated into the
test, but the “unknown unknown” can never be included. We do the
best with what we can imagine.

The situation is different when the tester wants only to determine

which of a finite number of pre-defined possible perceptions is
being controlled, as in Rick’s demos. Which is the car whose
location you are controlling, of three possibilities? The answer is
quite clear, without trying to answer exactly what defines the
perception of “location” that you are controlling.

Anway, the point is that for many purposes one can treat the

environmental correlate of the perceptual function as though it were
being controlled exactly as the corresponding perception is
controlled, but when you come to detail you have to be a bit
cautious about it, especially when the candidate perceptions are at
a high level, as Robertson et al. found when they used the TCV in a
study of self-image (Robertson, R. J., Goldstein, D. M., Mermel, M.,
& Musgrave, M. (1999). Testing the self as a control system:
Theoretical and methodological issues. International journal of
human-computer studies, 50, 571-580).

Martin

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

[MartinT]

[Martin Taylor 2015.03.23.14.45]

Which I endeavoured to make more precise, but you failed to comment

on my argument or on the point made by the argument.
After explaining why, my point was exactly as I said it…
Martin

···

On 2015/03/23 12:47 PM, Richard Marken
( via csgnet Mailing List) wrote:

rsmarken@gmail.com

[From Rick Marken (2015.03.23.0945)]

            Martin Taylor

(2015.03.23.09.44)–

---

                            [Fred

Nickols]***

---

                            So

behavior is also the control of aspects
of our environment? How does that jibe
with “We only control perception.�?***

            MT: Rick's statement might be less controversial if on

the third line he had added a couple of extra words, so
that the sentence would read “So when we control
perceptions, we seem to be controlling aspects of the
environment that correspond to those perceptions.”

          RM: This certainly does describe the phenomenology. We

experience our perceptions of the environment as the
environment. I, for example, experience the computer I’m
typing on as an actual thing out there in the environment,
not as a perception. It is a perception, of course, but I
don’t experience it that way.Â

          RM: But that wasn't my point. My point was exactly as I

said it:

            Â ...the point is

that for many purposes one can treat the environmental
correlate of the perceptual function as though it were
being controlled exactly as the corresponding perception
is controlled, but when you come to detail you have to
be a bit cautious about it, especially when the
candidate perceptions are at a high level, as Robertson
et al. found when they used the TCV in a study of
self-image (Robertson, R. J., Goldstein, D. M., Mermel,
M., & Musgrave, M. (1999). Testing the self as a
control system: Theoretical and methodological issues.
International journal of human-computer studies, 50,
571-580).

            Martin

[rupert]

[From Rupert Young (2015.03.23 21.30)]

(Rick Marken (2015.03.23.0945)]

Is it not the case that when we control perceptions we are always

controlling perceptions, but only sometimes controlling
aspects of the environment that correspond to those perceptions? For
example, if we want to perceive the alignment of two objects, trees
say, then we move ourself until we perceive one tree behind the
other. In this case there is no environmental correlate of the
perception, it only “exists” internally, so there is nothing in the
environment that we are controlling.

Regards,

Rupert

···

RM: But that wasn’t my point. My point was exactly as I
said it: When we control perceptions we are also
controlling aspects of the environment that correspond to
those perceptions. Perhaps to get my point across more
clearly I should have said: When we control perceptions we
are also controlling aspects of the environment (per the
current models of physics and chemistry) that correspond
to those perceptions.

[rsmarken]

[From Rick Marken (2015.03.23.1545)]

···

Martin Taylor (2015.03.23.14.45)–

MT: Which I endeavoured to make more precise, but you failed to comment

on my argument or on the point made by the argument.

MT: After explaining why, my point was exactly as I said it…

RM: I don’t see how this could be an exception. According to the PCT model, a perception, p, is always a function of environmental variables, x.1,x.2…x.n. That is. p = f(x.1,x.2…x.n). This is true whether p is a low level perception, like the angle of a limb, or a high level perception, like the honesty of a sales pitch. So whenever a control system controls a perception of any type, high level or low, it is also controlling the aspect of the environment, x.1,x.2…x.n, defined by f().

RM: The main difference I see between higher and lower level perceptions is that we know how to compute the function that produces a lower level perception, such as the angle of a limb, given the environmental variables on which it is based; but we can’t yet do that for a higher level perception, like the honesty of a sales pitch. So we can model control of lower level perceptual variables, like limb angle, because we can simulate the function, f(), that produces those perceptions using a computer. But we won’t be able to model control of higher level perceptions, like honesty of a sales pitch, until we figure out how to simulate the function that produces a perception honesty from the environmental inputs – tone of voice, content of what is said, postural “tells”, etc – on which it is based. So for the time being the study of control of higher level variables will have to use the perceptual systems of humans to derive the perceptions that might be under control.

RM: But,again, according to the PCT model, all perceptions are functions, ultimately, of environmental variables. So, in principle, if a machine (or another human) can compute a function of environmental variables that is the same as the one a control system is using to “construct” the perception it is controlling then that machine (or person) is “looking at” at the perception that the control system is controlling.

Best

          RM: But that wasn't my point. My point was exactly as I

said it:

            MT:  ...the point is

that for many purposes one can treat the environmental
correlate of the perceptual function as though it were
being controlled exactly as the corresponding perception
is controlled, but when you come to detail you have to
be a bit cautious about it, especially when the
candidate perceptions are at a high level, as Robertson
et al. found when they used the TCV in a study of
self-image (Robertson, R. J., Goldstein, D. M., Mermel,
M., & Musgrave, M. (1999). Testing the self as a
control system: Theoretical and methodological issues.
International journal of human-computer studies, 50,
571-580).

Rick

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

[rsmarken]

[From Rick Marken (2015.03.23.1615]

···

Rupert Young (2015.03.23 21.30)

    (Rick Marken (2015.03.23.0945)]

RY: Is it not the case that when we control perceptions we are always

controlling perceptions, but only sometimes controlling
aspects of the environment that correspond to those perceptions? For
example, if we want to perceive the alignment of two objects, trees
say, then we move ourself until we perceive one tree behind the
other. In this case there is no environmental correlate of the
perception, it only “exists” internally, so there is nothing in the
environment that we are controlling.

RM: Great example. I’ve tried to be careful to say that it is an aspect of the environment that is controlled when you control a perception, not necessarily the environment itself. In your example, the perception controlled is the optical angle between the trees. This could be controlled by actually operating on the environment and moving the trees. But it is far simpler to control this variable by, as you say,moving sideways until the optical angle between the trees is 0. By doing this you are controlling an aspect of your environment, which is the physical relationship between you and the trees.

RM: Perceptions always exist only internally. But there is always an environmental correlate to a perception. In your example, the environmental correlate is the physical location of the trees and your position relative to them. The perception of the angular difference between the trees would be different if the position of the trees (and you relative to them) were different. Also, the fact that there is an environmental correlate of the perception is demonstrated by the fact that another person can have the same perception as you had when you visually aligned the trees if the person moves relative to the same trees from the same distance you did --and if they are able to perceptive optical angles as you did. The person will not have the same perception as you had if they stand farther back or closer in to the trees.

RM: I think it’s hard to deny that there is an environment out there made up of variables from which our perceptions are constructed by our perceptual functions. So when we control perceptions we are always controlling an aspect of the environment on which those perceptions depend. Because of this it is possible for an outside observer, using the principles of the test for the controlled variable, to determine what perception(s) someone else is controlling.

Best

Rick

–

          RM: But that wasn't my point. My point was exactly as I

said it: When we control perceptions we are also
controlling aspects of the environment that correspond to
those perceptions. Perhaps to get my point across more
clearly I should have said: When we control perceptions we
are also controlling aspects of the environment (per the
current models of physics and chemistry) that correspond
to those perceptions.

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

[DagArchive]

RM: … According to the PCT
model, a perception, p, � is always a function of environmental
variables, x.1,x.2…x.n. That is. p = f(x.1,x.2…x.n). This is true
whether p is a low level perception, like the angle of a limb, or a high
level perception, like the honesty of a sales pitch.
[From Dag Forssell (2015 19:30 PDT)]

···

At 03:47 PM 3/23/2015, you wrote:

Well, the PCT model does not account for memory. But when you enjoy your
beloved Bach, memory is very much involved. You anticipate every note,
don’t you.

So perceptions are functions of environvental variables, plus a huge
amount retreived from stored memory.

The PCT model is limited in scope this way. That is OK.

Time to change thread heading ???

Best, Dag

[FredNickols]

[From Fred Nickols (2015.03.24.0521 EDT)]

Hmm. If memory comes into play in a conscious way wouldn’t that be a perception?

Fred Nickols

···

From: Dag Forssell (csgarchive@pctresources.com via csgnet Mailing List) [mailto:csgnet@lists.illinois.edu]
Sent: Monday, March 23, 2015 10:34 PM
To: csgnet@lists.illinois.edu
Subject: Re: FW: Control Theory Entry in APA Dictionary of Psychology

[From Dag Forssell (2015 19:30 PDT)]

At 03:47 PM 3/23/2015, you wrote:

RM: … According to the PCT model, a perception, p, Â is always a function of environmental variables, x.1,x.2…x.n. That is. p = f(x.1,x.2…x.n). This is true whether p is a low level perception, like the angle of a limb, or a high level perception, like the honesty of a sales pitch.

Well, the PCT model does not account for memory. But when you enjoy your beloved Bach, memory is very much involved. You anticipate every note, don’t you.

So perceptions are functions of environvental variables, plus a huge amount retreived from stored memory.

The PCT model is limited in scope this way. That is OK.

Time to change thread heading ???

Best, Dag

[MartinT]

[Martin Taylor 2015.03.24.23.30]

I suppose I have to recopy my entire argument, since you answer as

though you hadn’t looked at it, despite my request that you comment
on the argument.
To which I might now add that the arguments of P( ) are not only
inevitably delayed as compared to the current state of the
environment, but more generally include elements of history – the
prior state of that perception and its input variables, which, as
Dag pointed out include specific memory, and as “”.
That’s a mighty big “If”, since the machine would have to have had
the same perceptual constructions at all levels over the entire life
history of the system being tested, and be subject to the same
modulating influences as the subject for its result to be exact. The
TCV is great for getting an approximation to what is being
controlled when the possible variation is continuous, and even
better for finding which of a discrete set of possibilities is now
being controlled, such as “for which of these cars is the subject
controlling some aspect of on-screen location?” And those results
can be very useful. But one should never believe that the TCV could
determine exactly what perception is being controlled in any real
situation.
That wasn’t the point of my reference to Robertson et al… Their
issue was one of timing. They didn’t get very good results in the
first couple of studies, and they hypothesised that the problem was
that the perception of self-image was continually changing, so that
the TCV had to include a disturbance as close in time to the error
correction as they could manage. When they did that, they were able
to claim that the self-image did seem to be (poorly) controlled by
some of the subjects, but the best that they could get out of it was
that the question needed further study. The TCV, in other words,
showed that something related to what they hypothesised was being
controlled by most subjects, but not what, and not whether it was
the same thing from subject to subject.
David Goldstein was a co-author on the paper, and it would be
interesting to know whether anything further has been done in the
intervening decade and a half.
Martin

···

On 2015/03/23 6:47 PM, Richard Marken
( via csgnet Mailing List) wrote:

rsmarken@gmail.com

[From Rick Marken (2015.03.23.1545)]

            Martin Taylor

(2015.03.23.14.45)–

                        RM: But that wasn't my point. My point

was exactly as I said it:

            MT: Which I endeavoured to make more precise, but

you failed to comment on my argument or on the point
made by the argument.

            MT: After explaining why, my point was exactly as I said

it…

                        MT:

…the point is that for many purposes one
can treat the environmental correlate of the
perceptual function as though it were being
controlled exactly as the corresponding
perception is controlled, but when you come
to detail you have to be a bit cautious
about it, especially when the candidate
perceptions are at a high level, as
Robertson et al. found when they used the
TCV in a study of self-image (Robertson, R.
J., Goldstein, D. M., Mermel, M., &
Musgrave, M. (1999). Testing the self as a
control system: Theoretical and
methodological issues. International journal
of human-computer studies, 50, 571-580).

          RM: I don't see how this could be an exception.

According to the PCT model, a perception, p, is always a
function of environmental variables, x.1,x.2…x.n. That
is. p = f(x.1,x.2…x.n). This is true whether p is a low
level perception, like the angle of a limb, or a high
level perception, like the honesty of a sales pitch. So
whenever a control system controls a perception of any
type, high level or low, it is also controlling the aspect
of the environment, x.1,x.2…x.n, defined by f().

[Martin Taylor 2015.03.23.09.44]
…the perceptual function defines how
environmental changes influence the perceptual value. If
perception = P(environment), then if P( ) is monotonic,
environment = P^-1(perception).

  Because of the inevitable time delay between a change in the

environment and its influence on the perception, if P( ) is
causal, then P^-1( ) is non-causal, which means non-physical. The
perception does not cause anything to happen in the environment,
but this does not matter, because we are only trying to determine
what perception is being controlled. If the time-lag between
environmental change and the corresponding change in perception is
small compared to the loop transport lag, we can treat the
environmental function as being equivalent to the perception, and
therefore as being controlled, even though it is technically not
being controlled.

  The only access we have to the perceptions others may be

controlling is the “Test for the Controlled Variable” (TCV). The
TCV depends (as it must) on effects in the world observable by the
tester. The tester can never have exactly the same perceptions of
the world as the subject of the Test, but in many cases,
especially for low-level perceptions, the two sets of perceptions
can be assumed to be pretty close. So, when testing a parametric
range of possible candidates for the controlled perception, the
one that results in the appearance of best control of the
hypothesised environmental function is the best bet for a
definition of the perceptual function. Of course, it is always
possible that what is actually being controlled includes some
variable not incorporated into the test, but the “unknown unknown”
can never be included. We do the best with what we can imagine.

  The situation is different when the tester wants only to determine

which of a finite number of pre-defined possible perceptions is
being controlled, as in Rick’s demos. Which is the car whose
location you are controlling, of three possibilities? The answer
is quite clear, without trying to answer exactly what defines the
perception of “location” that you are controlling.
[From: Richard Pfau
(2015.03.24 1645 EST)] pointed out, are modulated by * other
internal factors, mechanisms, and related descriptive
concepts include hormones, drugs, expectations, overall affect
such as mood, etc*

          RM: But,again, according to the PCT model, all

perceptions are functions, ultimately, of environmental
variables. So, in principle, if a machine (or another
human) can compute a function of environmental variables
that is the same as the one a control system is using to
“construct” the perception it is controlling then that
machine (or person) is “looking at” at the perception that
the control system is controlling.

    [Martin Taylor

2015.03.23.09.44][Anway, the point is that for many purposes one
can treat the environmental correlate of the perceptual function
as though it were being controlled exactly as the corresponding
perception is controlled, but when you come to detail you have
to be a bit cautious about it, especially when the candidate
perceptions are at a high level, as Robertson et al. found when
they used the TCV in a study of self-image (Robertson, R. J.,
Goldstein, D. M., Mermel, M., & Musgrave, M. (1999). Testing
the self as a control system: Theoretical and methodological
issues. International journal of human-computer studies, 50,
571-580).

  [From Rick Marken (2015.03.23.1545)]: The

main difference I see between higher and lower level perceptions
is that we know how to compute the function that produces a lower
level perception, such as the angle of a limb, given the
environmental variables on which it is based; but we can’t yet do
that for a higher level perception, like the honesty of a sales
pitch. So we can model control of lower level perceptual
variables, like limb angle, because we can simulate the function,
f(), that produces those perceptions using a computer. But we
won’t be able to model control of higher level perceptions,
like honesty of a sales pitch, until we figure out how to simulate
the function that produces a perception honesty from the
environmental inputs – tone of voice, content of what is said,
postural “tells”, etc – on which it is based. So for the time
being the study of control of higher level variables will have to
use the perceptual systems of humans to derive the perceptions
that might be under control.

[rsmarken]

[From Rick Marken (2015.03.25.0955)]

···

Dag Forssell (2015 19:30 PDT)–

RM: … According to the PCT
model, a perception, p, Â is always a function of environmental
variables, x.1,x.2…x.n. That is. p = f(x.1,x.2…x.n). This is true
whether p is a low level perception, like the angle of a limb, or a high
level perception, like the honesty of a sales pitch.
DF: Well, the PCT model does not account for memory.

RM: I think you must mean that the PCT model doesn’t currently include memory as one of the inputs to the perceptual function. There is a PCT account of memory in the “Memory” chapter of B:CP (Ch 15 in both editions). I think Bill didn’t include memory as an input to the perceptual function because, as far as I know, there is no evidence that suggests that it should be.

DF: But when you enjoy your
beloved Bach, memory is very much involved. You anticipate every note,
don’t you.

RM: Yes, memory is very much involved but not as an input to a perception but as a reference for the perception. The “anticipation” that I experience when listening to Bach is, I believe, a reference for what I should be perceiving, not an input to my perception itself. And that’s where memory is in the PCT model of memory; in the reference signals (see Fig. 15.3, p 223). Memory starts with a reference signal that serves as the “address” signal that selects stored reference signals to be sent to lower level systems. When those reference signals are played back into the perceptual signal path of the system that selected the address for the reference we have a memory – not a perception but an imagined perception corresponding to the stored reference .

RM: The imagined perception could be occurring while the actual perception is occurring; but I usually am not imagining what I am hearing while I’m hearing it. But I do have a reference for what I should be hearing. So I think I’m listening to music in passive observation mode (p. 222); the perception of the music is being compared to the reference for what I should be hearing but if there is a discrepancy between perception and reference – an error – I do nothing about it because the output “switch” of the control loop is not connected in passive observation mode. Memory is involved in this only inasmuch as it provides the reference against which the musical perception is being compared.

RM: I had an interesting experience while listening to the 3rd movement of the Brahms violin concerto on the radio a while ago that is consistent with this model. Towards the end of the third and final movement I heard that something was wrong; I couldn’t put my finger on it but there was definitely a difference between what I expected to hear (had a reference for hearing) and what I heard (the perception of the concerto). It turns out that it was a recording of a live concert and the DJ announced that the violinist, Anna Sophie Mutter, broke a sting shortly before the finale and the first violinist had smoothly handed his violin to Anna in mid measure. But a few notes had been missed and that resulted in the error signal for me. It turns out that Anna Sophie would only let that recording be played in public if it was accompanied by the announcement of the broken string after the recording had completed.

RM: What’s interesting is that the “error” could be detected only by someone who remembered the music pretty well. I was with my wife, who is not nearly as familiar with the Brahms violin concerto as I am, and she was astounded that I heard the “error”. Indeed, when I noticed the “error” I said to her “something happened; something went wrong”. Then when the DJ announced the broken strong she was very impressed. Which just shows that even in passive observation mode you can impress people by detecting errors, even if you can’t correct them yourself;-)

DF: So perceptions are functions of environvental variables, plus a huge
amount retreived from stored memory.

RM: I won’t believe it until I see some data that demands an explanation using a model where perception includes a huge amount retrieved from stored memory. My subjective experience is that my perceptions, like my perception of the Brahms violin concerto, contain no memory at all; but my evaluation of my perceptions is done with respect to my remembered reference specifications for what those perceptions should be.

Best

Rick

The PCT model is limited in scope this way. That is OK.

Time to change thread heading ???

Best, Dag

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

[rsmarken]

[From Rick Marken (2015.03.25.1130)]

Martin Taylor (2015.03.24.23.30)--

RM: I don't see how this [higher level perceptions] could be an exception. According to the PCT model, a perception, p, is always a function of environmental variables, x.1,x.2...x.n. T

MT: I suppose I have to recopy my entire argument, since you answer as though you hadn't looked at it, despite my request that you comment on the argument.

Martin Taylor (2015.03.23.09.44)
MT: ...the perceptual function defines how environmental changes influence the perceptual value. If perception = P(environment), then if P( ) is monotonic, environment = P^-1(perception).

RM: For me, the "environment" is the models of physics and chemistry. Using these models we try different perceptual functions, P(environment), to see which gives us the best fit to the observed behavior. We never go from perception to environment, nor do we imagine that we can; the environment is a _theory; all we have that is not theoretical is our perception. Therefore, I'm puzzled by your claim that you can get the environment by taking the inverse of the perceptual function of a perception, P^-1(perception). What would be the environment you get from taking the inverse of the perceptual function that produces "green", say? There is actually more than one set of (theoretical) environmental variables -- wavelengths of light -- that produces "green". Given that I perceive green how do I get from there to the wavelengths that are causing that perception?

MT: Because of the inevitable time delay between a change in the environment and its influence on the perception, if P( ) is causal, then P^-1( ) is non-causal, which means non-physical. The perception does not _cause_ anything to happen in the environment, but this does not matter, because we are only trying to determine what perception is being controlled. If the time-lag between environmental change and the corresponding change in perception is small compared to the loop transport lag, we can treat the environmental function as being equivalent to the perception, and therefore as being controlled, even though it is technically not being controlled.

RM: I'm sorry, I just don't see what transport lag has to do with getting a good estimate of the perception that is under control. There is evidence that the transport lag is greater for higher level than for lower level perceptions but all these perceptions, high level and low, can be controlled so it's pretty easy to figure out what they are by applying disturbances and seeing whether or not they are effective (see Marken, R. S., Khatib, Z. and Mansell, W. (2013) Motor Control as the Control of Perception, Perceptual and Motor Skills, 117, 236-247).

MT: The only access we have to the perceptions others may be controlling is the "Test for the Controlled Variable" (TCV). The TCV depends (as it must) on effects in the world observable by the tester. The tester can never have exactly the same perceptions of the world as the subject of the Test, but in many cases, especially for low-level perceptions, the two sets of perceptions can be assumed to be pretty close. So, when testing a parametric range of possible candidates for the controlled perception, the one that results in the appearance of best control of the hypothesised environmental function is the best bet for a definition of the perceptual function. Of course, it is always possible that what is actually being controlled includes some variable not incorporated into the test, but the "unknown unknown" can never be included. We do the best with what we can imagine.

RM: Yes, absolutely.

MT: The situation is different when the tester wants only to determine which of a finite number of pre-defined possible perceptions is being controlled, as in Rick's demos. Which is the car whose location you are controlling, of three possibilities? The answer is quite clear, without trying to answer exactly what defines the perception of "location" that you are controlling.

MT: To which I might now add that the arguments of P( ) are not only inevitably delayed as compared to the current state of the environment,

RM: As I said, this is irrelevant to the success of the test.

MT: but more generally include elements of history -- the prior state of that perception and its input variables, which, as Dag pointed out include specific memory, and as [From: Richard Pfau (2015.03.24 1645 EST)] pointed out, are modulated by "other internal factors, mechanisms, and related descriptive concepts include hormones, drugs, expectations, overall affect such as mood, etc".

RM: I trust data more than people "pointing things out". I am not aware of any data demonstrating the need to include memory in perceptions. I think there is pretty good evidence that drugs can affect perceptions but I am not sure that any of the other things that you list actually affect perception (as opposed, perhaps, to the evaluation of perception). >>

RM: But,again, according to the PCT model, all perceptions are functions, ultimately, of environmental variables. So, in principle, if a machine (or another human) can compute a function of environmental variables that is the same as the one a control system is using to "construct" the perception it is controlling then that machine (or person) is "looking at" at the perception that the control system is controlling.

MT: That's a mighty big "If", since the machine would have to have had the same perceptual constructions at all levels over the entire life history of the system being tested, and be subject to the same modulating influences as the subject for its result to be exact.The TCV is great for getting an approximation to what is being controlled when the possible variation is continuous, and even better for finding which of a discrete set of possibilities is now being controlled, such as "for which of these cars is the subject controlling some aspect of on-screen location?" And those results can be very useful. But one should never believe that the TCV could determine _exactly_ what perception is being controlled in any real situation.

RM:OK, you can't determine exactly what the controlled perceptions are but it's better than not having any idea what the controlled perceptions are or, even worse, not even knowing that behavior is organized around controlled perceptions (as is the case in current research in psychology).

RM: The main difference I see between higher and lower level perceptions is that we know how to compute the function that produces a lower level perception, such as the angle of a limb, given the environmental variables on which it is based; but we can't yet do that for a higher level perception, like the honesty of a sales pitch.

MT: That wasn't the point of my reference to Robertson et al.. Their issue was one of timing.

RM: Yes, I re-read their paper. It was an interesting study but don't think timing had anything to do with the failure of the first try. I think in the first experiment there was no consistent resistant to "disturbance" because those disturbance were not disturbances to anything the subject was controlling. In the second and third experiment the subjects had a clearer idea of what they were to control so the "disturbance" words were countered consistently. That's the way the test works; if you are not getting disturbance resistance then your disturbances are not disturbances, meaning that the person is not controlling what you thought they were controlling. Tie for a new hypothesis about the controlled variable, which is basically what Robertson et al did.>

MT: David Goldstein was a co-author on the paper, and it would be interesting to know whether anything further has been done in the intervening decade and a half.

RM: I agree.
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&gt;Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

[MartinT]

[Martin Taylor 2015.03.25.15.41]

You don't. I'm mystified as yo how this relates to the statement

that “if P( ) is monotonic, environment = P^-1(perception)”. We
aren’t asking about the perception of wavelength or mixtures of
wavelength. We are asking about the perception of “green”, which is
a label. Labels apply to whole categories, and clearly are not
monotonic invertible functions of anything.

In my statement, I was trying to provide you with some support when

you talk about things in the environment being controlled, by
pointing out that there do exist situations in which this can be
said to be true.

      No. What it has to do with

is whether the perceptual lag need be taken into account in the
context of control.

What is the evaluation of perception if it is not a perception? I

suspect that what you are saying is that you are not aware of any
data showing low-level perceptions based directly on sensory data to
be dependent on anything other than sensory data. But go up a
little, and consider my oft-used example of someone pressing a
button beside a door. As an onlooker, you may use sensory data alone
to perceive the act of pushing a small domey thing. But you need
memory to perceive John calling on his Aunt Mabel, or even to
perceive a man ringing a doorbell.

We are in violent agreement her!

Martin

···

[From Rick Marken (2015.03.25.1130)]

            Martin Taylor

(2015.03.24.23.30)–

            MT: I suppose I have to recopy my entire

argument, since you answer as though you hadn’t looked
at it, despite my request that you comment on the
argument.

Martin Taylor (2015.03.23.09.44)
MT: …the perceptual function
defines how environmental changes influence the
perceptual value. If perception = P(environment), then
if P( ) is monotonic, environment = P^-1(perception).

          RM: For me, the "environment" is the models of physics

and chemistry. Using these models we try different
perceptual functions, P(environment), to see which gives
us the best fit to the observed behavior. We never go from
perception to environment, nor do we imagine that we can;
the environment is a _theory; all we have that is not
theoretical is our perception. Therefore, I’m puzzled by
your claim that you can get the environment by taking the
inverse of the perceptual function of a perception,
P^-1(perception). What would be the environment you get
from taking the inverse of the perceptual function that
produces “green”, say? There is actually more than one set
of (theoretical) environmental variables – wavelengths of
light – that produces “green”. Given that I perceive
green how do I get from there to the wavelengths that are
causing that perception?

                        RM: I don't see how this [higher level

perceptions] could be an exception.
According to the PCT model, a perception, p,
is always a function of environmental
variables, x.1,x.2…x.n. T

                MT: Because of the

inevitable time delay between a change in the
environment and its influence on the perception, if
P( ) is causal, then P^-1( ) is non-causal, which
means non-physical. The perception does not cause
anything to happen in the environment, but this does
not matter, because we are only trying to determine
what perception is being controlled. If the time-lag
between environmental change and the corresponding
change in perception is small compared to the loop
transport lag, we can treat the environmental
function as being equivalent to the perception, and
therefore as being controlled, even though it is
technically not being controlled.

            RM: I'm

sorry, I just don’t see what transport lag has to do
with getting a good estimate of the perception that is
under control.

          RM: I trust data more than people "pointing things

out". I am not aware of any data demonstrating the need to
include memory in perceptions. I think there is pretty
good evidence that drugs can affect perceptions but I am
not sure that any of the other things that you list
actually affect perception (as opposed, perhaps, to the
evaluation of perception).

            MT: That's a mighty big "If", since the machine

would have to have had the same perceptual constructions
at all levels over the entire life history of the system
being tested, and be subject to the same modulating
influences as the subject for its result to be exact.The
TCV is great for getting an approximation to what is
being controlled when the possible variation is
continuous, and even better for finding which of a
discrete set of possibilities is now being controlled,
such as “for which of these cars is the subject
controlling some aspect of on-screen location?” And
those results can be very useful. But one should never
believe that the TCV could determine exactly what
perception is being controlled in any real situation.

          RM:OK, you can't determine exactly what the controlled

perceptions are but it’s better than not having any idea
what the controlled perceptions are or, even worse, not
even knowing that behavior is organized around controlled
perceptions (as is the case in current research in
psychology).

                        RM: But,again, according to the PCT

model, all perceptions are functions,
ultimately, of environmental variables. So,
in principle, if a machine (or another
human) can compute a function of
environmental variables that is the same as
the one a control system is using to
“construct” the perception it is controlling
then that machine (or person) is “looking
at” at the perception that the control
system is controlling.