Applied and Basic PCT research (was Re: Expression of emotions)

[Martin Taylor 2017.06.22.12.53]

[From Rick Marken (2017.06.21.1630)]

Not within a PCT framework, according to which every intentional

action is for control of some variable.

That there IS an interaction, in which each party is trying to

discover the intentions of the other – at several levels, I should
add. They do this by acting in ways that would disturb the
perceptions they hypothesise that the other is controlling, and
observe the actions produced by the other.

Oh, well, I just pointed out another kind.

I wouldn't expect to, since the proportion of people engaged in

cooperative or non-cooperative interactions who have heard of the
TCV is probably close to one in a billion. If they haven’t heard of
it, they would not perceive themselves as performing it. But that
doesn’t mean that they aren’t.

I hereby ask Warren.

But you did hint at the result, so I presume you have the relevant

data, and that’s the interesting part.

[From Rick Marken (2017.05.29.1930)]

    Martin

Taylor (2017.05.28.07.14)–

    MT:

Rick Marken frequently claims that the only legitimate research
object of PCT is the search for the controlled variable.

    RM: Not quite. What I claim is that the main focus of

research based on PCT has to be testing to determine what
variables are controlled when we see organisms carrying out
various behaviors.This was Bill Powers’ vision for a research
program based on PCT; it is the program he describes in many
places but most explicitly in his paper “A Cybernetic Model for
Research in Human Development”, reprinted in Living Control
Systems. The reason for this is that the PCT model posits that
the observable behavior of organisms is organized around the
control of a hierarchy of perceptual variables. Powers has
hypothesized, based on subjective experience, that there are at
least nine levels of this hierarchy, each level consisting of
control systems controlling a different type of
perceptual variable.

    RM: This is the central feature of the PCT model and,

therefore, it is the aspect of the model that should receive the
greatest amount of attention from research, just as it was the
aspect of the model that received the greatest amount of
attention (in terms of pages of discussion) in B:CP. But there
are certainly other “legitimate” areas for PCT-based research.
There has already been some nice research on reorganization
(Robertson and Glines) and it would be good to see more of that.
It would also be nice to see some research testing the PCT
models of memory and imagination. But I think the focus of PCT
research should be on the determination of the variables that
organisms control.

Which is a little different from

[From Rick Marken (2017.06.21.1630)]

  RM: ... But all PCT research is centered around understanding what

perceptions organisms control and how they control them.

It's the use of "all", which frequently occurs in this context, that

causes the friction. Your earlier quote is much more reasonable.
People can reasonably have different opinions as to where they want
to focus their own research into PCT, and where they expect the most
important results to come from.

As in all kinds of research, there's a continuum between basic and

applied research. Basic has very general applicability but needs
specific boundary conditions and parameters if it is to be used in
any specific situation, whereas applied starts with the boundary
conditions and parameters and has little or no generalizability.
Applied solves a particular problem for an immediate purpose,
whereas basic is like exploration, providing information that might
later prove helpful for some applied problem, or on which later
research can build further.

What a particular person controls in a particular situation is "hard

applied" PCT research, whereas studying the effects of transport lag
on controllability against arbitrary disturbances is “pretty hard
basic” PCT research. In between there are studies such as “What
types of perceptual variables are controlled by predators but not by
prey, and why?” I see no basis, other than (as Rick has told us)
Bill P’s preference, for saying any one kind of research should be
considered primary over any other. I see applied and basic research
as “keeping each other honest”.

I'll give a personal example from another domain entirely, effective

colouring of satellite imagery.

Some time around 1969, I proposed that a theoretically based

approach to colouring the then new Landsat I imagery would allow
interpreters to see much clearer distinctions among different kinds
of ground cover, and mocked up a kind of illustration of what I
expected to deliver. I thought that it might take about 3 weeks to
complete the work. Instead, I found that in practice there were many
things I had not thought of, such as variable atmospheric haze,
solar lighting angle, and so on and so on. It too two or three years
to produce useful results (published in 1973). Applied research can
show up missing elements in basic research, and often does.

But my idea depended on basic research that had shown (to

oversimplify greatly) that colour perception was based the three
basic cone receptor types R, G, and B in the following way:
Brightness = R+B+G, Red-Green balance = R-G, and Blue-Yellow balance
= (R+G)-B, in that order of importance. Up to that time, the
four-band Landsat imagery had been displayed by choosing three of
the bands and showing near infra-red as red, and two of the others
(I forget which) as blue and green. My idea was to do a principal
components analysis of the four bands and display the first
component as brightness, the second as red-green balance, and the
third as blue-yellow balance (where “yellow” was R+G). The final
result was pretty good, as this example of an area of Ellesmere
Island in the high Arctic may suggest. The left panel is the then
standard method of display, while the right is a principal
components presentation of the same data (notice particularly the
area on the left side and the streak from top right to bottom left,
which are indistinguishable in the “standard” display.

![Ellesmere.PCA.Demo.jpg|673x306](upload://pZ48lFrm56SEWXVoU5XlZ3siA3F.jpeg)

A contractor actually commercialized this process in hardware

offered for sale to the satellite imagery community around 1974. In
1983 it was reinvented at JPL and that reinvention was cited in IEEE
“Spectrum” as a great advance in imagery display.

I think this example is a good one for illustrating the interplay

between basic and applied research. The core idea came from basic
research, but it took a good deal of work to deal with the practical
issues that usually are associated with applied research.

The same, I think, applies to different aspects of PCT research.

Bill was able to use basic concepts and demonstrations analogous to
my first mock-ups, along with his own introspection, to create the
hierarchy. The specific hierarchic levels have not been tested other
than by saying that the “seem reasonable” (despite that even at the
lowest levels, Bill used to try out changing the order a little). Is
the real hierarchy in an organism actually a hierarchy with no
lateral connections? We don’t know from PCT-based research, though
other research suggests that it probably is not. Is the “category
level” a level at all? Perhaps not. Applied research is probably
required to test these and other possibilities.

To apply PCT research in a specific context requires appreciable

work (searching for controlled variables, with mock-up
demonstrations that the search is often feasible, and hard-earned
results in real specific situations). To apply the basic research in
more complex situations requires more basic research in, for
example, the many ways in which control systems can interact.
Applied research in example complex situations is required if the
basic research is not to go too far into Cloudland, and basic
research is necessary for laying the foundations for applied tests
and the solution of particular problems.

So I fully support that the search for the controlled variable(s) is

of continuing importance, but not that it is of primary importance
in the development of PCT.

Martin
···

Martin Taylor (2017.06.20.23.30)–

            MT: I'm glad you

haven’t. Incidentally, I’ve been writing that
cooperative interaction (and some uncooperative
interaction) between and among individuals can be seen
as a mutual use of the TCV on each other.

          RM: It can also be viewed as a stimulus-response

sequence.

          What is the evidence that these interactions involve

mutual use of the TCV?

          The only evidence I could imagine is data showing that

the members of the cooperative interaction are controlling
for doing the TCV.

How you would obtain such evidence, I don’t know,

          but it would be an example of using the TCV to

determine whether people are controlling for a program
perception; the program “carrying out the TCV”. That is,
it would be using the TCV to test for control of the TCV.
That would be a fantastic demonstration if you could do
it!

            MT: That's something I didn't know, or if I did,

I had forgotten. I’d like to know more… Was it
published?

RM: I think so. You’d have to ask Warren.

                      RM: and I did it in modeling the results of

some studies of personal space (done by
students of Warren Mansell), where the space
maintained between two people was derived from
data on the space maintained by each of those
people in interaction with others.

[From Rick Marken (2017.06.23.1845)]

Ellesmere.PCA.Demo.jpg

···

[Martin Taylor 2017.06.22.12.53]Â

MT: That there IS an interaction, in which each party is trying to

discover the intentions of the other – at several levels, I should
add. They do this by acting in ways that would disturb the
perceptions they hypothesise that the other is controlling, and
observe the actions produced by the other.

RM: The fact that there is an interaction between people is not evidence that the parties to the interaction are carrying out the TCV. Indeed, I consider it highly unlikely that they are carrying out the TCV. But maybe they are. It has to be tested!

MT: Oh, well, I just pointed out another kind.

RM: As I said, the fact that people are interacting is not evidence that they are carrying out the TCV.Â

MT: I wouldn't expect to, since the proportion of people engaged in

cooperative or non-cooperative interactions who have heard of the
TCV is probably close to one in a billion. If they haven’t heard of
it, they would not perceive themselves as performing it. But that
doesn’t mean that they aren’t.

RM: Your hypothesis is that people who are interacting are carrying out the TCV. The TCV is a program of actions inasmuch as it involves at least one contingency (if-then branch). So you are hypothesizing that the participants in an interaction are controlling a program perception called the TCV that can be described as follows:Â

  1. Hypothesize a new controlled variable

  2. Apply disturbance that should have an effect on the variable if it is not controlled

  3. If the disturbance does have an effect the goto 1

  4. Else tentatively assume the hypothesized variable is controlled

RM: This is a simplified version of the program that might plausibly be controlled in an interaction. It would be tough to test for control of this program in a natural interaction. But there are probably ways to test it by having people interact with a conversational computer program. But, as I said, I don’t know, off hand, how to design a test for control of such a program perception. But it would be an interesting challenge to design it. And it would be a great demonstration of the PCT approach to understanding behavior. I did design a very simple test of control of a program perception. The results are described briefly in my “Hierarchical Behavior of Perception” paper on pp. 100-102 of More Mind Readings. It was a version of the “Hierarchical Behavior of Perception” demo (http://www.mindreadings.com/ControlDemo/Hierarchy.html). In the on-line version I only demonstrate control of configuration, transition and sequence perceptions. I will develop an on-line version of control of a program perception as soon as I get a chance.Â

MT: I hereby ask Warren.

MT: But you did hint at the result, so I presume you have the relevant

data, and that’s the interesting part.

RM: Yes, I have the relevant data. And the modeling that was used to predict it. The results were pretty good but I think they could have been much better if the data were collected a bit differently.Â

    RM: Â But I think the focus of PCT

research should be on the determination of the variables that
organisms control.Â

MT: Which is a little different from

  RM: ... But all PCT research is centered around understanding what

perceptions organisms control and how they control them.
MT: It’s the use of “all”, which frequently occurs in this context, that
causes the friction.

RM: Actually, I’m willing to say just that all PCT research, whatever it’s aim, should specify the variable(s) that are assumed to be controlled. So I agree that all PCT research – that is, all research that recognizes that behavior is a process of control of input – need not be aimed at determining what perceptions are controlled, But it should always be cognizant of the fact that the research either assumes or aims determine that some variable is controlled!

RM: Your treatise below on applied and basic research actually made me realize that this is the case. (Congratulations on your clever method of displaying Landsat imagery). It reminded me that I had done some applied research that didn’t use the TCV per se. It was my research on the possible causes of prescription (Rx) errors. Actually I published a couple papers on this; the most “applied” one is here:

https://www.dropbox.com/s/5z6mswgmxdk5dmr/ModelPrioritize.pdf?dl=0

RM: You’ll see that I simply assumed that the controlled variable was a correctly written Rx. I did fit the model to some error data, for different components of an Rx. So actually, there were four controlled variables assumed, corresponding to the four components of an Rx for which there was error data.Â

RM: Anyway, whether it’s applied or basic research, I’d say that if you’re not at least mentioning possible controlled variables, you are not doing PCT research.Â

BestÂ

Rick

Â

Your earlier quote is much more reasonable.

People can reasonably have different opinions as to where they want
to focus their own research into PCT, and where they expect the most
important results to come from.

As in all kinds of research, there's a continuum between basic and

applied research. Basic has very general applicability but needs
specific boundary conditions and parameters if it is to be used in
any specific situation, whereas applied starts with the boundary
conditions and parameters and has little or no generalizability.
Applied solves a particular problem for an immediate purpose,
whereas basic is like exploration, providing information that might
later prove helpful for some applied problem, or on which later
research can build further.

What a particular person controls in a particular situation is "hard

applied" PCT research, whereas studying the effects of transport lag
on controllability against arbitrary disturbances is “pretty hard
basic” PCT research. In between there are studies such as “What
types of perceptual variables are controlled by predators but not by
prey, and why?” I see no basis, other than (as Rick has told us)
Bill P’s preference, for saying any one kind of research should be
considered primary over any other. I see applied and basic research
as “keeping each other honest”.

I'll give a personal example from another domain entirely, effective

colouring of satellite imagery.

Some time around 1969, I proposed that a theoretically based

approach to colouring the then new Landsat I imagery would allow
interpreters to see much clearer distinctions among different kinds
of ground cover, and mocked up a kind of illustration of what I
expected to deliver. I thought that it might take about 3 weeks to
complete the work. Instead, I found that in practice there were many
things I had not thought of, such as variable atmospheric haze,
solar lighting angle, and so on and so on. It too two or three years
to produce useful results (published in 1973). Applied research can
show up missing elements in basic research, and often does.

But my idea depended on basic research that had shown (to

oversimplify greatly) that colour perception was based the three
basic cone receptor types R, G, and B in the following way:
Brightness = R+B+G, Red-Green balance = R-G, and Blue-Yellow balance
= (R+G)-B, in that order of importance. Up to that time, the
four-band Landsat imagery had been displayed by choosing three of
the bands and showing near infra-red as red, and two of the others
(I forget which) as blue and green. My idea was to do a principal
components analysis of the four bands and display the first
component as brightness, the second as red-green balance, and the
third as blue-yellow balance (where “yellow” was R+G). The final
result was pretty good, as this example of an area of Ellesmere
Island in the high Arctic may suggest. The left panel is the then
standard method of display, while the right is a principal
components presentation of the same data (notice particularly the
area on the left side and the streak from top right to bottom left,
which are indistinguishable in the “standard” display.

A contractor actually commercialized this process in hardware

offered for sale to the satellite imagery community around 1974. In
1983 it was reinvented at JPL and that reinvention was cited in IEEE
“Spectrum” as a great advance in imagery display.

I think this example is a good one for illustrating the interplay

between basic and applied research. The core idea came from basic
research, but it took a good deal of work to deal with the practical
issues that usually are associated with applied research.

The same, I think, applies to different aspects of PCT research.

Bill was able to use basic concepts and demonstrations analogous to
my first mock-ups, along with his own introspection, to create the
hierarchy. The specific hierarchic levels have not been tested other
than by saying that the “seem reasonable” (despite that even at the
lowest levels, Bill used to try out changing the order a little). Is
the real hierarchy in an organism actually a hierarchy with no
lateral connections? We don’t know from PCT-based research, though
other research suggests that it probably is not. Is the “category
level” a level at all? Perhaps not. Applied research is probably
required to test these and other possibilities.

To apply PCT research in a specific context requires appreciable

work (searching for controlled variables, with mock-up
demonstrations that the search is often feasible, and hard-earned
results in real specific situations). To apply the basic research in
more complex situations requires more basic research in, for
example, the many ways in which control systems can interact.
Applied research in example complex situations is required if the
basic research is not to go too far into Cloudland, and basic
research is necessary for laying the foundations for applied tests
and the solution of particular problems.

So I fully support that the search for the controlled variable(s) is

of continuing importance, but not that it is of primary importance
in the development of PCT.

Martin


Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

          RM: What is the evidence that these interactions involve

mutual use of the TCV?

          RM: The only evidence I could imagine is data showing that

the members of the cooperative interaction are controlling
for doing the TCV.

RM: How you would obtain such evidence, I don’t know,

            MT: That's something I didn't know, or if I did,

I had forgotten. I’d like to know more… Was it
published?

 RM: I think so. You’d have to ask Warren.

                      RM: and I did it in modeling the results of

some studies of personal space (done by
students of Warren Mansell), where the space
maintained between two people was derived from
data on the space maintained by each of those
people in interaction with others.