PCT enquiry

[rsmarken]

[From Rick Marken (2015.01.22.1600 PST)]

I received some questions about PCT from Angus Jenkinson who I believe is relatively new to PCT, although he is a colleague of James Wilk, who is a longtime fan of PCT. I made a personal reply to Angus but realized that this might benefit from going on CSGNet so I asked Angus if it was OK to move this to CSGNet and he said yes so here we go. I’ll try to repeat the relevant parts of the conversation up to this point, with AJ being Angus and RM being yours truly.

···

*AJ: I did not intend to suggest that PCT did not work in fact, or that it was merely theory, although of course it is called perceptual control theory. *

RM: As usual I  wasn’t clear. I didn’t mean that PCT works in fact. I meant that PCT only applies to behavior that is, in fact, purposeful: control behavior. So the behavior of a ball rolling down an inclined plane or that of a marble coming to rest at the bottom of a bowl does not involve control (as can be demonstrated using the Test for the Controlled Variable); these behaviors are not, in fact, purposeful. So PCT is not needed to explain these behaviors. But the behavior of a person throwing a ball or running down the side of a hill is, in fact, purposeful – it is control behavior. It can be shown that, in both cases, there are variables that are being controlled and PCT is required to explain how this control is effected.Â

Â

• AJ: I am however slightly surprised by some of your remarks. Although I
  know that Richard [William–RM] Powers was rooted in engineering control theory, they also regarded him as a cybernetician and the theory rooted in cybernetics. Much more significantly, it seems rather vital to PCT that it should work “all the Way downâ€?. *

RM: To the extent that it’s control “all the way down” then, yes, I would hope that PCT would work “all the way down”. That was the point of my saying that PCT applies only to behavior that demonstrably involves control. To the extent that control is involved in cellular processes, such as mitosis (and I suspect that it is) then PCT would apply at this level. I guess my point is simply that, from a PCT perspective, it’s not accounting for behavior at all levels of a living organism that matters; it’s accounting for the controlling done by living organisms, regardless of the level of the organism at which the controlling is done. PCT is explains that controlling done by living (and artifactual) systems.

Â

• AJ: A key aspect of
  PCT is that involves the cancellation of perturbation, which is crudely the inverse of conventional causality.*

RM: Yes, a control organization cancels the perturbing effect that variables (called disturbances) would have on a variable that is controlled by the organism. Perturbation cancellation is control; PCT explains how it works (control of a perceptual representation of the controlled variable in a closed negative feedback loop). Conventional causality works fine as an explanation of non-control behavior, such as the ball rolling down an inclined plane. The ball doesn’t act to cancel perturbations, such as a change in the slope of the place as it descends; it’s behavior is precisely predictable from the cause effect laws of physics. PCT is not needed to explain it’s behavior.

Â

• AJ: If the driver steering the car is using PCT (behaviour to control perceptions of the outside world), then would that not apply through the organism,
  e.g. to the wrists and the heartbeat? *

RM: The driver steering a car is controlling the position of the car on the road, keeping it in it’s lane. PCT is a model of how the driver does that: by acting to keep a perception of the car’s position relative to the lane in a reference state specified by the driver. The driver is, indeed, acting (behaving) to control a perception of a variable in the outside world. To the extent that wrist position and heart rate are under control and control of those variables is part of (or disturbed by) control of the driver’s position on the road then, yes, PCT also applies to that controlling. But, again, the “applicability” of PCT depends on whether or not the phenomenon to be explained involved control (and steering a car, moving one’s wrist and having a heartbeat are all behaviors that do involve control).Â

Â

*AJ: For example a heart responds to (the perturbation of) a flow of liquid, even when removed from the body. *

RM: I don’t know much about it but if the heart has flow sensors then I imagine that it could control the flow of liquid outside the body. The heart would then be not just a pump but a control system in itself.

Â

AJ: Moreover, if the temperature in the car is getting hot, he/she will also begin sweating to control body temperature

RM: Yes, so there is control involved and PCT would certianly be able to explain it.Â

Â

AJ: and this will involve an activity of the vagus nerve (cells) (as well as dilation of pores and many other things) and for all of this we are surely not going to revert back to traditional causality and its accompanying baggage?

RM: I think PCT would say that we should render unto causality what is causal and render unto PCT what is control. Control of body temperature is a control process and, therefore, can be explained only with a closed-loop control model: ie. PCT. But the reduction in the temperature of a sauce pan that results when I hold it udder cold water is a causal (not a control) process that can be explained by traditional causality.Â

Â

• AJ: Efferent and afferent nerves
  (efferent being a much better name than motor nerve) transfer signals between centre and periphery, much as Stafford Beer models in VSM, unsurprisingly, with feedback loops to maintain control? Or have I gone radically beyond the bounds of the theory?*

RM: Not at all. Indeed, I don’t think you have gone far enough! In PCT afferent and efferent neurons don’t just transfer signals between center and periphery; efferent neurons are seen as carrying signals that are specifications for the values of signals carried by afferent neurons. Efferent signals are central (and autonomously) set specifications for what should be experienced (perceived).Â

Â

AJ: That is rather as I understood it and I am glad to get the confirmation.

RM: Great!Â

• AJ: I am sure that the proving work you are doing is important in generating interest and credibility for the theory. Over time it will no doubt need to develop the hierarchical model and all kinds of other implications for biological processes (as discussed
  above). But we have been living for a long time with cognitive and other models that have also lacked a satisfactory hierarchical model.*

RM: Â Yes, I think developing a good hierarchical model of human organization has to be one of the main focuses of research on purposeful behavior. It would be nice to hear some ideas about the kind of research that could be done to flesh out such a model.Â

Best regards

Rick

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

AJ: There are two areas that intrigue me – which is also to say leave me uncomfortable with what I have found so far.

RM: Great! I think all ideas should be approached with skepticism!Â

AJ: The first of these, and from a practical point of view the lesser of the two in importance for me, is working through the implications of PCT through the entire organism. The scientist after scientist, what I see is a big bold idea that is then explained
in terms of the old paradigm (typical for example of complexity theory). So if we take the principle of PCT on board, it refers not simply to the human being but all the way down to every cell (at least). And this certainly has a occasions but we understand
the brain to be and the nature of distributed intelligence. I am wondering where you see the latest and best thinking in relationship to this?

RM: I’m not sure I understand the question. But I don’t think I agree with the premise: that the “principle” of PCT applies “all the way down to every cell”. I think the main contribution of PCT is the recognition that the behavior of organisms is a process
of control, in fact, not in theory. Control, which is also known as purposeful behavior, involves bringing variable aspects of an organism’s experience to pre-specified states and maintaining them there, protected from disturbance. PCT is an explanation of
how organisms do this controlling – how control works. PCT is simply engineering control theory properly applied to the controlling done by living systems.Â

RM: To the extent that cells control variables in their environment (and they probably do) then the PCT applies to the behavior of cells as well. But the application of PCT always depends on establishing that the behavior to be explained involves control.
So  PCT does not necessarily apply to the whole organism – nor does it claim to. It just applies to the phenomenon of control as it is seen in the behavior of living organisms. To the extent that components of those organisms (or groups to which those organisms
belong) behave purposefully – ie, control – then PCT applies. But PCT applies only to the extent that the phenomenon to be explained involved control; it doesn’t automatically apply to phenomena – such as the chemical processes involved in digestion –
simply because they occur in organisms.Â

AJ: And the second is concerned with the hierarchy. Once again I am concerned to make sure that the hierarchy is not based on old models.

RM: The hierarchy is a theoretical guess based on Bill Powers’ own introspection regarding the types of perceptions her was able to identify in his own experience and the relationship between those perceptions. It is meant to be a source of hypotheses for
research. I’m attaching a paper of mine that demonstrates the kind of research that might be done to test Bill’s hypothesis about the hierarchy.Â

RM: So the hierarchy is not based on old models or new ones; it’s based on Bill’;s subjective experience. It will almost certainly be changed as research progresses. But since there’s only a couple of people doing actual research based on PCT it will probably
be a long time before there is any real change. But the hierarchy is not etched in stone, by any means. Even the idea that the relationship between control systems is hierarchical is up for grabs. PCT science has just barely started!

AJ: Let me take for example the concept of system at the top of the hierarchy. Today that is an increasingly problematic word, because it has been hijacked by so much reductionism seen as a form of additive assembly of a whole, because there are so many different
forms of systems theory, and  (commonly) without actually explaining things in the way that cybernetics does. Powers’ proposal was seen as an hypothesis , an initial proposal, as I understand it, and I am wondering a) what kind of empirical work is being done
and B) what alternative models may have been developed

RM: I’ve attached one little paper on the topic. I don’t know of much other work on the topic. There are a number of hierarchical models described in Powers last book, LCS III, that could be considered alternatives to the original inasmuch as the variables
controlled at the differ net levels don’t necessarily match up with the type  (and hierarchical order) of the variables proposed in the original model. But there is really not much work on the hierarchy per se that I know of. Â

[rsmarken]

[From Rick Marken (2015.01.23.1025)]

···

Martin Taylor (2015.01.22.23.08)–

MT: Good answers, Rick, but I think two points need to be added.

RM: Thanks Martin!

MT: 1. Control versus homeostasis, both of which were mentioned, at least implicitly. Control differs from homeostasis, though both stabilize internal variables.

RM: I guess I don’t really understand the difference between control and homeostasis. I thought homeostasis was control. But the term “homeostasis” is mainly used to refer to control of physiological variables, like body temperature or glucose concentration. Is there really some fundamental difference between control and homeostasis? If so, could you show me how to write a program for a working “homeostasis” model so I can see how it differs from a working control model (which I already know to write).

MT: Homeostasis can function with just a negative feedback loop, whereas control needs a substantial asymmetry between input side and output side. With control, there is an inside and an outside, a reference value and a controlled value that is compared with the reference value.

RM: Isn’t that true of homeostasis as well? How can there be homeostasis, such as control of body temperature – without a comparison between controlled and reference value?

MT: 2. Causality versus control (and homeostasis)

Control and homeostasis are structural. Causal is simply a statement that if variation in one variable affects the value of another, the effect is later than the cause. Both control and homeostasis use causality, but only around a loop with certain specific characteristics. A simple causal connection can never control. It’s the structural relationship of causality around a loop that matters,

RM: Yes! Good point. And one that Bill often made, though instead of “structure” he would use the word “organization”. Control exists because of the organization of cause and effect, the organization that results in control being a closed negative feedback loop.

MT: so that an effect at any place in the loop is a part of the cause of later values at that place. Causality is everywhere, in the sense that the values of all the variables depend on the past and not on the future. No kind of structural relationship is implied by causality. Control can exist only where the specific causal relationships create an asymmetric negative feedback loop structure in which a controlled variable can be distinguished. Control is an emergent property of the structure, not of the causality.

RM: Righto!

Best

Rick

–

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

[MartinT]

[Martin Taylor 2014.01.23.13.35]

As I understand it, the essence of a control loop is the asymmetry

between the input side and the output side. Homeostasis can function
with any kind of negative feedback loop. I guess you could call it a
generalized kind of control loop if the loop parameters are
appropriate, but I think it is easier just to keep “control” for the
kind of loop in B:CP, rather than saying that some kinds of
homeostatic loops are control loops but if you change the gains a
bit, then they aren’t. Anyway, lets analyse this one:
The G’s are inverting multipliers so that, for example, p1 = -G1s1.
The loop gain is -G1G2G3 = -G
Let’s look at how p1 is affected by the disturbances d1, d2, and d3.
p1 = -G1s1
= -G1*(p3+d3)
= -G1*(-G3s3 + d3)
= -G1(-G3*(p2+d2) + d3)
= -G1*(-G3*(-G2s2 + d2) + d3)
= -G1(-G3*(-G2*(p1+d1) +d2) +d3)
p1(1+G) = -Gd1 + d2(G/G2) - d3*(G/G2G3)
p1 = d1(-G/(1+G)) + d2*(G/G2*(1+G)) - d3*(G/G2G3(1+G))
The other p formulae are similar, just rotating the subscripts.
p1 relates to d1 the way p in a control loop relates to the
reference value. If G >> 1, p1 ~ d1 (~ means “is approximately
equal to”)
The effect of the other d values decreases the further around the
loop is the connection.
To simplify, lets say that all the Gn are equal to g = G^-3 (cube
root of G)
p1 = d1*(-g^3/(1+G)) + d2*(g^2/(1+G)) - d3*(g/(1+G))
If, say, d1 is a constant, then p1 relates to d2 as perception in a
control loop relates to its reference, but with a lower loop gain.
Globally, the loop values are influenced by the disturbances, but
unequally. The longer the loop, the less is the effect of
immediately preceding disturbances, and the more nearly does the p
value at any stage follow the next incoming d value. In other words,
all around the loop you have p values stabilized at the local
reference value, while the other d inputs act somewhere between
references and disturbances, depending on the relative gains.
The whole loop isn’t stabilized, any more than is the case in a
standard control loop, but any point in the loop can be.
You can, of course, extend this to a loop with more than three
stages, or reduce it to two, the normal control loop if G1>>G2
(and you make sure there is an odd number of sign inversions around
the loop).
Martin

···

[From Rick Marken (2015.01.23.1025)]

          MT: 1. Control versus homeostasis, both of which were

mentioned, at least implicitly. Control differs from
homeostasis, though both stabilize internal variables.

          RM: I guess I don't really understand the difference

between control and homeostasis.

[rsmarken]

Hi Martin

Angus said he wasn’t getting out replies so then I noticed that you have sent this only to me so I’m copying you replies to CSGNet and to Angus just in case.

···

On Thu, Jan 22, 2015 at 8:29 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[From Rick Marken (2015.01.22.1600 PST)]

I received some questions about PCT from Angus Jenkinson who I believe is relatively new to PCT, although he is a colleague of James Wilk, who is a longtime fan of PCT. I made a personal reply to Angus but realized that this might benefit from going on CSGNet so I asked Angus if it was OK to move this to CSGNet and he said yes so here we go. I’ll try to repeat the relevant parts of the conversation up to this point, with AJ being Angus and RM being yours truly.
[Martin Taylor 2015.01.22.23.08]

Good answers, Rick, but I think two points need to be added.

1. Control versus homeostasis, both of which were mentioned, at least implicitly.

Control differs from homeostasis, though both stabilize internal variables. Homeostasis can function with just a negative feedback loop, whereas control needs a substantial asymmetry between input side and output side. With control, there is an inside and an outside, a reference value and a controlled value that is compared with the reference value.

2. Causality versus control (and homeostasis)

Control and homeostasis are structural. Causal is simply a statement that if variation in one variable affects the value of another, the effect is later than the cause. Both control and homeostasis use causality, but only around a loop with certain specific characteristics. A simple causal connection can never control. It’s the structural relationship of causality around a loop that matters, so that an effect at any place in the loop is a part of the cause of later values at that place. Causality is everywhere, in the sense that the values of all the variables depend on the past and not on the future. No kind of structural relationship is implied by causality. Control can exist only where the specific causal relationships create an asymmetric negative feedback loop structure in which a controlled variable can be distinguished. Control is an emergent property of the structure, not of the causality.

Martin

–

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

[PHILIP_JERAIR_YERANO]

Could someone please remind me what sort of vacuum tube amplifier Bill likened his negate feedback pct loops to.

[rsmarken]

[From Rick Marken (2014.01.26.1600)]

···

Martin Taylor (2014.01.23.13.35)_-

MT: As I understand it, the essence of a control loop is the asymmetry

between the input side and the output side. Homeostasis can function
with any kind of negative feedback loop…

RM: And control can function only in what kind of negative feedback loop?

MT: Let’s look at how p1 is affected by the disturbances d1, d2, and d3.

p1 = -G1*s1

     = -G1*(p3+d3)

     = -G1*(-G3*s3 + d3)

     = -G1*(-G3*(p2+d2) + d3)

     = -G1*(-G3*(-G2*s2 + d2) + d3)

     = -G1*(-G3*(-G2*(p1+d1) +d2) +d3)

p1(1+G) = -G*d1 + d2*(G/G2) - d3*(G/G2*G3)

RM: I don’t understand how you got to that last equation. Where did G come from?

p1 = d1*(-G/(1+G)) + d2*(G/G2*(1+G)) - d3*(G/G2*G3*(1+G))

RM: I think there are some subscripts off but I get the gist.

MT: The other p formulae are similar, just rotating the subscripts.

p1 relates to d1 the way p in a control loop relates to the

reference value. If G >> 1, p1 ~ d1 (~ means “is approximately
equal to”)

RM: So all the d’s are like possible reference values?

MT: The effect of the other d values decreases the further around the

loop is the connection.

To simplify, lets say that all the Gn are equal to g = G^-3 (cube

root of G)

p1 = d1*(-g^3/(1+G)) + d2*(g^2/(1+G)) - d3*(g/(1+G))



If, say, d1 is a constant, then p1 relates to d2 as perception in a

control loop relates to its reference, but with a lower loop gain.

RM: What about d3? Doesn’t it get to function as a reference also?

MT: Globally, the loop values are influenced by the disturbances, but

unequally. The longer the loop, the less is the effect of
immediately preceding disturbances, and the more nearly does the p
value at any stage follow the next incoming d value. In other words,
all around the loop you have p values stabilized at the local
reference value, while the other d inputs act somewhere between
references and disturbances, depending on the relative gains.

RM: So there are some temporal relationships going on here. I’m not sure I’m following your analysis. But I’ll take your word for it.

RM: Actually, what would really help me understand this is if you could explain how to distinguish the behavior of a homeostatic system from that of a control system.

Best

Rick

The whole loop isn't stabilized, any more than is the case in a

standard control loop, but any point in the loop can be.

You can, of course, extend this to a loop with more than three

stages, or reduce it to two, the normal control loop if G1>>G2
(and you make sure there is an odd number of sign inversions around
the loop).

Martin

–

          RM: I guess I don't really understand the difference

between control and homeostasis.

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

[MartinT]

[Martin Taylor 2015.01.26.23.02]

The kind described by Bill P in B:CP. I suppose there may be other

kinds, but that’s what I have always thought of as a “control loop”,
in which a perceptual value is controlled by acting through an
environment to keep the perceptual value near a possibly changing
reference value in the face of variable disturbances.
From the line immediately above the part you quoted, reproduced here
with the accompanying figure.
Wherein are subscripts off?
And like possible disturbances at the same time.
As a reference for p3, yes. As does d2 for p2.
There are, of course, but I didn’t mention them because you usually
like the static algebraic analysis. I just did it your usual way,
ignoring all the temporal questions. But I’m glad you do see that
dynamics matters, even though I carefully did not mention it in my
message.
Below your signature, we find the following quote from my previous
message.
Martin

···

[From Rick Marken (2014.01.26.1600)]

            Martin Taylor

(2014.01.23.13.35)_-

                        RM: I guess I don't really understand the

difference between control and homeostasis.

            MT: As I understand it, the essence of a control loop is

the asymmetry between the input side and the output
side. Homeostasis can function with any kind of negative
feedback loop…

          RM: And control can function only in what kind of

negative feedback loop?

            MT: Let's look at

how p1 is affected by the disturbances d1, d2, and d3.

            p1 = -G1*s1

                 = -G1*(p3+d3)

                 = -G1*(-G3*s3 + d3)

                 = -G1*(-G3*(p2+d2) + d3)

                 = -G1*(-G3*(-G2*s2 + d2) + d3)

                 = -G1*(-G3*(-G2*(p1+d1) +d2) +d3)

            p1(1+G) = -G*d1 + d2*(G/G2) - d3*(G/G2*G3)

          RM: I don't understand how you got to that last

equation. Where did G come from?

[Martin Taylor 2014.01.23.13.35]

  The G's are inverting multipliers so that, for example, p1 =

-G1s1. The loop gain is -G1G2*G3 = -G

            p1 = d1*(-G/(1+G)) + d2*(G/G2*(1+G)) -

d3*(G/G2G3(1+G))

          RM: I think there are some subscripts off but I get the

gist.

            MT: The other p

formulae are similar, just rotating the subscripts.

            p1 relates to d1 the way p in a control loop relates to

the reference value. If G >> 1, p1 ~ d1 (~ means
“is approximately equal to”)

RM: So all the d’s are like possible reference values?

            MT: The effect of the other d values decreases the

further around the loop is the connection.

            To simplify, lets say that all the Gn are equal to g =

G^-3 (cube root of G)

            p1 = d1*(-g^3/(1+G)) + d2*(g^2/(1+G)) - d3*(g/(1+G))



            If, say, d1 is a constant, then p1 relates to d2 as

perception in a control loop relates to its reference,
but with a lower loop gain.

          RM: What about d3? Doesn't it get to function as a

reference also?

            MT: Globally, the

loop values are influenced by the disturbances, but
unequally. The longer the loop, the less is the effect
of immediately preceding disturbances, and the more
nearly does the p value at any stage follow the next
incoming d value. In other words, all around the loop
you have p values stabilized at the local reference
value, while the other d inputs act somewhere between
references and disturbances, depending on the relative
gains.

          RM: So there are some temporal relationships going on

here. I’m not sure I’m following your analysis. But I’ll
take your word for it.

          RM: Actually, what would really help me understand this

is if you could explain how to distinguish the behavior of
a homeostatic system from that of a control system.

            The whole loop isn't stabilized, any more than is the

case in a standard control loop, but any point in the
loop can be.

            You can, of course, extend this to a loop with more than

three stages, or reduce it to two, the normal control
loop if G1>>G2 (and you make sure there is an odd
number of sign inversions around the loop).

                Martin

[rsmarken]

[From Rick Marken (2015.01.27.0940)]

···

Martin Taylor (2015.01.26.23.02)–

MT: Below  your signature, we find the following quote from my previous

message.

RM: I don’t see how this allows be to distinguish the behavior of a homeostatic system from that of a control system. What I would like to have is an empirical test that would allow me to determine whether the behavior I am observing is homeostasis or control. Can you tell me how to do that?

          RM: Actually, what would really help me understand this

is if you could explain how to distinguish the behavior of
a homeostatic system from that of a control system.

            MT: The whole loop isn't stabilized, any more than is the

case in a standard control loop, but any point in the
loop can be.

            You can, of course, extend this to a loop with more than

three stages, or reduce it to two, the normal control
loop if G1>>G2 (and you make sure there is an odd
number of sign inversions around the loop).

Best

Rick

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

[Angus_Jenkinson]

Thank you Richard.

I am assuming that if I reply to CSGNet, my reply will be posted.

There are various tracks on the conversation to do with homeostat and control that I will sideline except to pose the following comments:

I think there is a relationship between both of these and identity; if we consider self reference and self organisation (pace Varela), then the ability to maintain (a word that neatly bisects homeostat and control) identity (in the sense) gives a larger context for any particular behaviour controlling perception.

I think when we are talking about organisms and in particular humans, we are dealing with intention and goals and that in some kind of hierarc
hy (assuming that is the right form to consider it). I am not convinced that the term reference is the best way to describe this as it reeks of a traditional cognitive model.

I take on board the distinction between phenomena that involve control and those that do not (see comment below). However, some of the conversation seems to assume causality (In the organism, the person, i.e. In controlled situations) when an essential feature of PCT is the negation of causality.

Responding to the earlier conversation…

• AJ: I am however slightly surprised by some of your remarks. Although I know that William Powers was rooted in engineering control theory, they also regarded him as a cyberneticia
  n and the theory rooted in cybernetics. Much more significantly, it seems rather vital to PCT that it should work “all the Way down”. *

RM: To the extent that it’s control “all the way down” then, yes, I would hope that PCT would work “all the way down”. That was the point of my saying that PCT applies only to behavior that demonstrably involves control. To the extent that control is involved in cellular processes, such as mitosis (and I suspect that it is) then PCT would apply at this level. I guess my point is simply that, from a PCT perspective, it’s not accounting for behavior at all levels of a living organism that matters; it’s accounting for the controlling done by living organisms, regardless of the level of the organism at which the controlling is done. PCT is explains that controlling done by living (and artifactual) systems.

AJ Thank you for confirming this.

RM: PCT is a model of how the driver does that: by acting to keep a perception of the car’s position relative to the lane in a reference state specified by the driver.

AJ: Are you familiar with the work of Dreyfus? I have a concern about “reference states”, already articulated above. Reference states have been a vital component of cognitive theory. Dreyfus proposes “skilful coping” which constitutes action that would enable a driver to see and make adjustments that were required (consciously or unconsciously) in order to maintain her path on the road, but without needing a reference state external to the activity. It seems intuitive that there must be some kind of reference state, but it also seems to be intuitive to some people tha
t the brain controls the whole process, that my actions are caused, that perception is passive and indeed the whole apparatus PCT contradicts.

*AJ: and this will involve an activity of the vagus nerve (cells) (as well as dilation of pores and many other things) and for all of this we are surely not going to revert back to traditional causality and its accompanying baggage? *

RM: I think PCT would say that we should render unto causality what is causal and render unto PCT what is control. Control of body temperature is a control process and, therefore, can be explained only with a closed-loop control model: ie. PCT. But the reduction in the temperature of a sauce pan that results when
I hold it udder cold water is a causal (not a control) process that can be explained by traditional causality.

Of course, fully agreed. Provided we limit the phenomenon under observation to the saucepan of water and not to you holding it, and therein we get a most interesting question when we come to larger ecologies and so forth. Heart valves for example “are not alive”. They are composed of collagen that is “inert” and under causal influence (in this model). But I suspect that as we begin to better understand the entangled nature of organism and larger ecologies/ecosystems this may need revisiting.

AJ: Efferent and afferent nerves (efferent being a much better name than motor nerve) transfer signals between centre and periphery, much as Stafford Beer models in VSM, unsurprisingly, with feedback loops to maintain control? Or have I gone radically beyond the bounds of the theory?

RM: Not at all. Indeed, I don’t think you have gone far enough! In PCT afferent and efferent neurons don’t just transfer signals between center and periphery; efferent neurons are seen as carrying signals that are specifications for the values of signals carried by afferent neurons. Efferent signals are central (and autonomously) set specifications for what should be experienced (perceived).

AJ Interesting, thank you.

So my understanding of this is that efferent neurons do not operate as motor neurons were assumed to do, i.e. the
y do not directly activate, effect action, motorise, make things happen; rather they provide information that a local autonomous perception-controlling agent uses?

What about the locally held reference (leaving aside my concern about reference above, for the moment)?

RM: I don’t know much about it but if the heart has flow sensors then I imagine that it could control the flow of liquid outside the body. The heart would then be not just a pump but a control system in itself.

AJ: The heart is most definitely not just a pump. There is evidence that it sets a variety of rhythmical resonances throughout the body as well.

…………̷
0;……………………………………….

[rsmarken]

[From Rick Marken (2015.01.27.1110)]

···

Martin Taylor (2014.01.27.12.44)–

Martin sent this just to me by accident so I’m copying the relevant parts and sending it to CSGNet.

MT: Can you tell me an empirical test to show whether the thing I am

looking at is a flower OR a rose?

RM: No, but I can tell you how to tell a rose from other types of flowers. If homeostasis is a type of control in the same way that a rose is a type of flower then there must be other types of control as well besides homeostasis. If homeostasis were the only kind of control then there would be no need to call it anything other than “control” just as there would be no need to call a rose anything other than “flower” if a rose were the only kind of flower. So what I need is a way to distinguish homeostasis from the other types of control that are not homeostasis.

Best

Rick

–

      RM: I don't see how this allows be to distinguish the behavior

of a homeostatic system from that of a control system. What I
would like to have is an empirical test that would
allow me to determine whether the behavior I am observing is
homeostasis or control. Can you tell me how to do that?

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

[rsmarken]

[From Rick Marken (2015.01.27.1240)]

···

On Tue, Jan 27, 2015 at 10:41 AM, Angus Jenkinson angus.jenkinson@me.com wrote:

Thank you Richard.

I am assuming that if I reply to CSGNet, my reply will be posted. Â

RM: Hi Angus. You made it! To help us keep track of who is posting and when they post we have adopted the convention of putting a heading at the top of the post (like mine abve) that give the name of the person posting and when they posted (YYYY,MM,DD).

Â

There are various tracks on the conversation to do with homeostat and control that I will sideline except to pose the following comments:

RM: Good idea. And I’ll just comment on the parts of your post that I think I can answer. If I miss anything that you think I (or anyone else) should comment on just let us know.Â

AJ: I take on board the distinction between phenomena that involve control and those that do not (see comment below). However, some of the conversation seems to assume causality (In the organism, the person, i.e. In controlled situations) when an essential feature of PCT is the negation of causality.Â

RM: I think it’s better to say that PCT is the negation of lineal (or “open-loop”) causality as the basis of behavior. There is still causality in PCT, it just runs in a circle: closed-loop causality. Â

RM: PCT is a model of how the driver does that: by acting to keep a perception of the car’s position relative to the lane in a reference state specified by the driver.

AJ: Â Are you familiar with the work of Dreyfus?Â

RM: Only Julia Louise Dreyfus from “Seinfeld” and the guy who got railroaded in France back in the early 1900s.Â

Â

AJ: I have a concern about “reference statesâ€?, already articulated above. Reference states have been a vital component of cognitive theory.Â

RM: Not in the way they exist in PCT, I don’t believe.Â

Â

AJ: Dreyfus proposes “skilful coping� which constitutes action that would enable a driver to see and make adjustments that were required (consciously or unconsciously) in order to maintain  her path on the road, but without needing a reference state external to the activity.

RM: I don’t believe that it is possible to build a model of driving (control of car’s location on the road, for example)) that doesn’t include, at least implicitly, a reference specification for the state of the controlled variable (distance between car and lane). If Dreyfus has such a model then I would love to see it.Â

Â

AJ: It seems intuitive that there must be some kind of reference state, but it also seems to be intuitive to some people that the brain controls the whole process, that my actions are caused, that perception is passive and indeed the whole apparatus PCT contradicts.

RM: In PCT, reference (goal) states are specified  by efferent neural signals (reference signals) in the brain. The brain is part of the control process but it doesn’t really control the whole process. Control is itself a process and the role of the brain is to monitor the state of a controlled variable (perception), specify the desired state of that variable (reference signal) and vary outputs (actions) as the means of keeping the controlled variable in the desired state.

RM: Â In PCT afferent and efferent neurons don’t just transfer signals between center and periphery; efferent neurons are seen as carrying signals that are specifications for the values of signals carried by afferent neurons. Efferent signals are central (and autonomously) set specifications for what should be experienced (perceived).Â

AJ : Interesting, thank you.

So my understanding of this is that efferent neurons do not operate as motor neurons were assumed to do, i.e. they do not directly activate, effect action, motorise, make things happen; rather they provide information that a local autonomous perception-controlling agent uses?

RM: At the lowest level of the neural hierarchy of control efferent neurons directly cause action (by varying muscle length and tension). The lowest level efferents behave as neurophysiologists assume. So, for example, the motor neurons involved in spinal “reflexes”, for example, directly cause the action component of the reflex. But the efferents at the next levels up are specifications for afferent input.Â

Â

AJ: What about the locally held reference (leaving aside my concern about reference above, for the moment)?

RM: I don’t know what a “locally held reference” is. Perhaps you mean the lowest level efferents that connect directly to muscles. If so, as I said above, those efferents don’t function as reference signals. They are actually error signals that drive muscle contractions that affect perceptions (of tendon stretch, for example) that are under control in the loop.Â

Best regards

Rick

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

[MartinT]

[Martin Taylor 2014.01.27.14.27]

A rose is a type of flower. A flower is not a type of rose. Control

is a type of homeostasis. Homeostasis is not a type of control. Bruce Abbott has mentioned a type of homeostasis that is analogous
to a spring or a ball-in-a-bowl, in that the disturbance provides
the energy for the restoration of the status quo ante. Control
supplies the energy to oppose the disturbance independently of the
energy supplied by the disturbance. I have previously shown a
generic form of homeostasis involving negative feedback, and told
you the conditions under which this is the standard form of control.
I don’t like repeating myself, so this time I will ask you to do
your own analysis.
If you look at the diagram I presented, now presented for the third
time, maybe you can do your own static analysis instead of saying
you can’t follow mine.
When you have done that, try to see what the analysis shows if you
set G3 = -1, d2 = 0, -G1 >> -G3, G1G3 >> 0. Remember,
the gains Gn are simple multipliers, an odd number of them with
negative gain, to ensure that the loop gain G = G1G2*G3 is
negative. Then maybe you could look at a loop with other numbers of stages,
say 5, and see what general statements these easy analyses suggest
to you.
Martin

···

[From Rick Marken (2015.01.27.1110)]

            Martin Taylor

(2014.01.27.12.44)–

          Martin sent this just to me by accident so I'm copying

the relevant parts and sending it to CSGNet.

                    RM: I don't see how this

allows be to distinguish the behavior of a
homeostatic system from that of a control
system. What I would like to have is an empirical
test that would allow me to determine whether
the behavior I am observing is homeostasis or
control. Can you tell me how to do that?

            MT: Can you tell me an empirical test to show

whether the thing I am looking at is a flower OR a rose?

          RM: No, but I can tell you how to tell a rose from

other types of flowers. If homeostasis is a type of
control in the same way that a rose is a type of flower
then there must be other types of control as well besides
homeostasis. If homeostasis were the only kind of control
then there would be no need to call it anything other than
“control” just as there would be no need to call a rose
anything other than “flower” if a rose were the only kind
of flower. So what I need is a way to distinguish
homeostasis from the other types of control that are not
homeostasis.

[Angus_Jenkinson]

Apologies if I am butting into a conversation.

Control: the power to influence or direct… the ability to manage… the restriction of… a person or thing used as a standard…

Homeostasis: the tendency towards a relatively stable equili
brium between interdependent elements, especially as maintained by physiological processes.

Control, in this contex
t is the ability to maintain homeostasis?

Homeostasis requires control?

A homeostatic system is under control. A control system (in this context) maintains homeostasis.

They are internal and co-dependent to each other.

However there are forms of control (as generally used) not intending to produce homeostasis (equilibrium). Or requiring a revisioning of homeosta
sis.

Morphosis, or metamorphosis are not well described by homeostasis are they? Ontogeny. Although all these require multiple equilibria in the process and outcome

The artistic process (output not all the bodily behaviours), say the work of Kandinsky. The artist controls behaviours to bring about the perceptual goal (in a feedback loop), and their may be multiple centers and equilibria in the artwork, but homeostasis would be a limiting perspective surely?

Building a house.

Writing software.

Understanding a sentence or paragraph.

All require PCT

On the reverse perspective.

Seeing that the meaning of this sentence requires some form of interaction between meanings in context which settle into a meaning (“equilibrium point”) might give an interesting speculative way of expa
nding that notion?

Cheers

Angus

···

Martin Taylor (2014.01.27.12.44)–

Martin sent this just to me by accident so I’m copying the relevant parts and sending it to CSGNet.

MT: Can you tell me an empirical test to show whether the thing I am

looking at is a flower OR a rose?

RM: No, but I can tell you how to tell a rose from other types of flowers. If homeostasis is a type of control in the same way that a rose is a type of flower then there must be other types of control as well besides homeostasis. If homeostasis were the only kind of control then there would be no need to call it anything other than “control” just as there would be no need to call a rose anything other than “flower” if a rose were the only kind of flower. So what I need is a way to distinguish homeostasis from the other types of control that are not homeostasis.

Best

Rick

–

      RM: I don't see how this allows be to distinguish the behavior

of a homeostatic system from that of a control system. What I
would like to have is an empirical test that would
allow me to determine whether the behavior I am observing is
homeostasis or control. Can you tell me how to do that?

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

[Angus_Jenkinson]

2015,01,27 Angus Jenkinson

Thank you Richard

Closed-loop causality is causality. Surely?

I don’t see PCT as closed-loop causality.

I see PCT as active agency cancelling causality.

Help me to understand what causal factor loops back to ‘cause’ action as opposed to say occasion action or form a context in which an agent makes an action.

AJ: I have a concern about “reference states”, already articulated above. Reference states have been a vital component of cognitive theory.

RM: Not in the way they exist in PCT, I don’t believe.

• In other parts of your reply you state that "In PCT, reference (goal) states are specified by efferent neural signals (reference signals) in the brain.” Presumably this is proposing that the brain has an arrangement of neurons that constitutes a control (model) or reference for comparison. It computes a difference [or is there some other idea of the brain here?] and sends an instruction to correct the d
  ifference. Have I understood the idea? And if so how is this significantly different from cognitivism, other than the ordering of events? Is there any acceptance of the possibility of distributed local intelligence? How do habits like a pianist playing without being able to consciously intervene except at the musical level work?*

AJ: Dreyfus proposes “skilful coping” which constitutes action that would enable a driver to see and make adjustments that were required (consciously o
r unconsciously) in order to maintain her path on the road, but without needing a reference state external to the activity.

RM: I don’t believe that it is possible to build a model of driving (control of car’s location on the road, for example)) that doesn’t include, at least implicitly, a reference specification for the state of the controlled variable (distance between car and lane). If Dreyfus has such a model then I would love to see it.

Dreyfus gets rid of models.

R
M: At the lowest level of the neural hierarchy of control efferent neurons directly cause action (by varying muscle length and tension). The lowest level efferents behave as neurophysiologists assume. So, for example, the motor neurons involved in spinal “reflexes”, for example, directly cause the action component of the reflex. But the efferents at the next levels up are specifications for afferent input.

AJ: What about the locally held reference (leaving aside my concern about reference above, for the moment)?

RM: I don’t know what a “locally held reference” is. Perhaps you mean the lowest level efferents that connect directly to muscles. If so, as I said above, those efferents don’t function as reference signals. They are actually error signals that drive muscle contractions that affect perceptions (of tendon stretch, for example) that are under control in the loop.

Thank you.

*When the lowest level signal from the efferent neuron reaches the muscle cell why does the muscle cell(s) not cancel the signal? *

W* hen the signal (at any level) reaches the level below, why does the level below do wh
atever it does?*

Date: Tuesday, 27 January 2015 20:41

[From Rick Marken (2015.01.27.1240)]

···

On Tue, Jan 27, 2015 at 10:41 AM, Angus Jenkinson angus.jenkinson@me.com wrote:

Thank you Richard.

I am assuming that if I reply to CSGNet, my reply will be posted.

<
/div>

RM: Hi Angus. You made it! To help us keep track of who is posting and when they post we have adopted the convention of putting a heading at the top of the post (like mine abve) that give the name of the person posting and when they posted (YYYY,MM,DD).

There are various tracks on the conversation to do with homeostat and control that I will sideline except to pose the following comments:

RM: Good idea. And I’ll just comment on the parts of your post that I think I can answer. If I miss anything that you think I (or anyone else) should comment on just let us know.

AJ: I take on board the distinction between phenomena that involve control and those that do not (see comment below). However, some of the conversation seems to assume causality (In the organism, the person, i.e. In controlled situations) when an essential feature of PCT is the negation of causality.

RM: I think it’s better to say that PCT is the negation of lineal (or “open-loop”) causality as the basis of behavior. There is still causality in PCT, it just runs in a circle: closed-loop causality.

RM: PCT
is a model of how the driver does that: by acting to keep a perception of the car’s position relative to the lane in a reference state specified by the driver.

AJ: Are you familiar with the work of Dreyfus?

RM: Only Julia Louise Dreyfus from “Seinfeld” and the guy who got railroaded in France back in the early 1900s.

AJ: I have a concern about “reference states”, already articulated above. Reference states have been a vital component of cognitive theory.

<
div>RM: Not in the way they exist in PCT, I don’t believe.

AJ: Dreyfus proposes “skilful coping” which constitutes action that would enable a driver to see and make adjustments that were required (consciously or unconsciously) in order to maintain her path on the road, but without needing a reference state external to the activity.

RM: I don’t believe that it is possible to build a model of driving (control of car’s location on the road, for example)) that doesn’t include, at least implicitly, a reference specification for the state of the controlled variable (distance between car and lane). If Dreyfus has such a model then I would love to see it.

AJ: It seems intuitive that there must be some kind of reference state, but it also seems to be intuitive to some people that the brain controls the whole process, that my actions are caused, that perception is passive and indeed the whole apparatus PCT contradicts.

RM: In PCT, reference (goal) states are specified by efferent neural signals (reference signals) in the brain. The brain is part of the control process but it doesn’t really control the whole process. Control is itself a process and the role of the brain is to monitor the state of a controlled variable (perception), specify th
e desired state of that variable (reference signal) and vary outputs (actions) as the means of keeping the controlled variable in the desired state.

RM: In PCT afferent and efferent neurons don’t just transfer signals between center and periphery; efferent neurons are seen as carrying signals that are specifications for the values of signals carried by afferent neurons. Efferent signals are central (and autonomously) set specifications for what should be experienced (perceived).

AJ : Interesting, thank you.

So my understanding of this is that efferent neurons do not operate as motor neurons were assumed to do, i.e. they do not directly activate, effect action, motorise, make things happen; rather they provide information that a local autonomous perception-controlling agent uses?

RM: At the lowest level of the neural hierarchy of control efferent neurons directly cause action (by varying muscle length and tension). The lowest level efferents behave as neurophysiologists assume. So, for example, the motor neurons involved in spinal “reflexes”, for example, directly cause the action component of the reflex. But the efferents at the next levels up are specifications for afferent input.

AJ: What about the locally held reference (leaving aside my concern about referen
ce above, for the moment)?

RM: I don’t know what a “locally held reference” is. Perhaps you mean the lowest level efferents that connect directly to muscles. If so, as I said above, those efferents don’t function as reference signals. They are actually error signals that drive muscle contractions that affect perceptions (of tendon stretch, for example) that are under control in the loop.

Best regards

Rick

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

[rsmarken]

[From Rick Marken (2015.01.28.1720)]

2015,01,27 Angus Jenkinson
AJ: Help me to understand what causal factor loops back to ‘cause’ action as opposed to say occasion action or form a context in which an agent makes an action.

RM: In the simplest control loop -- such as the kind of control loop that results in control of a cursor in a tracking task (<http://www.mindreadings.com/ControlDemo/BasicTrack.html&gt;http://www.mindreadings.com/ControlDemo/BasicTrack.html\) there are two main causal components of the loop:
(1) o = k(r-p) and (2) p = o + d
RM: Equation 1 says that output (such as your mouse movements) is caused by the difference between perception, p (of cursor position, say) and the reference for the state of that perception, r (such as, "on target"). Equation 1 is the causal path from input (p) to output (o) via the organism. Equation 2 shows the causal path from output to input through the environment. In a tracking task this is the causal path from mouse movements to perceived cursor movements. Equation 2 also includes the effects of independent environmental disturbances, d, on the input.
RM: The two causal relationships described by equations 1 and 2 are occurring simultaneously so in order to see what kind of behavior is implied by these two equations we have to solve them simultaneously (note that p and o occur in both equations). When we solve these two equations for p and o we get
(3) p ~ r and (4) o ~ r - d
where "~" means nearly equal. Equation 3 says that the causal loop defined by equations 1 and 2 result in control of perception (the perceptual variable is brought into a match with the reference specification for what this perception should be) . And equation 4 says that the output of the control system will vary along with variations in the reference and, at the same time, in nearly perfect opposition to variations in the net disturbance, d, to the controlled variable, p.
RM: So there you have it; PCT in less than 300 words;-)

RM: I don't believe that it is possible to build a model of driving (control of car's location on the road, for example)) that doesn't include, at least implicitly, a reference specification for the state of the controlled variable (distance between car and lane). If Dreyfus has such a model then I would love to see it.

AJ: Dreyfus gets rid of models.

RM: Well then Dreyfus is of no interest to me because my understanding of control (purposive behavior) is all based on modeling. Indeed, modeling was central to Bill Powers' development and testing of PCT.

AJ: When the lowest level signal from the efferent neuron reaches the muscle cell why does the muscle cell(s) not cancel the signal?

RM: Because the muscle cell just responds to the efferent signal by lengthening or contracting.

When the signal (at any level) reaches the level below, why does the level below do wh atever it does?

RM: The function of an efferent at each level depends on the organization of the system at that level. If you have a copy of B:CP take a look at Figure 8.1 which shows two levels of control systems. The efferents at Level I (lowest level) go directly to the effectors which cause muscle actions that have an effect on variables in the outside world. The efferents at the next level up, Level II, enter the comparator functions of the control systems at Level I. They don't directly cause muscle action; rather they function as specifications for the Level I perceptions. Any difference between the Level I perceptions and the references for these perceptions (which are the efferents entering the Level I comparator from Level II) is an error signal in the form of the Level I efferent that directly drives action.
RM: I hope this helps.
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