Where Rick's Chapter 7 on "Social Control" goes off track

Jeez! Ever heard of the ‘imagination mode’?!

Hi Rick, yes, that last example is where we are getting at, and becomes particularly interesting if we think that different people inevitably (because their brains developed independently within their own craniums) have slightly different input functions for a variable, slightly different input signals from their perspective on the environment, and slightly different reference values for the variable, AND that control is hierarchical (the high level controlled variable is ‘car off snow’ but the lower level variables used to achieve this may be different (e.g. perceive car moving forwards; perceive snow melting; perceive a non-snowy path ahead), then collective control seems like the best way to describe it. Your summary is a handy simplification that is probably the most correct it could be as a simplification, but it doesn’t hold at a granular and multi-temporal level of analysis…

Substitute “example” for “case”. The point in this class of cases, of which this is one example, is that because not all participants perceive the variable concurrently, there is no conflict. This removes it from your fixation on conflict. You counter by denying that the state of a pothole in a public road is a collectively controlled variable. You deny this although it is perceived by many individuals in a public, any one or any plurality of whom may control their perceptions of it (swerving to avoid it, slowing to reduce the bump, calling Town Hall, stopping the patrol car and putting out a trestle with a blinking yellow light, consulting plans, schedules, and budgets, in one’s office far from the pothole, etc.).

It is too easy to suppose that this ‘pothole’ is a unitary homogeneous entity which is perceived in the same way by all who perceive it and who may seek to control such perceptions. It takes some memory of actual situations and some imagination of possible situations to adduce additional participants in the environing public, to see differences as to what aspects of a collectively controlled variable are perceived, and as to what aspects of the collectively controlled variable are affected by their diverse control capacities and processes. In an odd way, a pothole because of its seeming concrete specificity is too abstract an example—the many participants are not personally known to you and perhaps seem facelessly unknowable, and their diverse perceptual inputs, circumstances, and means of control are accordingly accessible only by imagination.

You object that in the example the two named participants have the same reference value, which you quantify as “zero”. (Odd looking asphalt patch, that zero.) But they are not the only participants, and the reference values of participants often differ for reasons including those that Warren adduced and because they have different uses of the given artifact in environmental feedback paths for their idiosyncratic control of other variables.

Suppose (heaven forefend) that your computer has a problem such that you can’t use it in numerous environmental feedback paths for controlling a great variety of variables which you are accustomed to control thereby. Let us stipulate (whether it is true or not) that you are not a computer technician. In this circumstance, you may well call on someone to fix it so let us further suppose that you do. She is also controlling numerous variables, some of them by environmental feedback paths that pass through your computer and its proper functioning or lack thereof. One of them might be a perception of you buying a new computer, but we’ll set that cynical thought aside. She is perceiving problems to be solved, money to be earned, creditworthiness, and many other variables which are to be controlled through her interventions in aspects of your computer that you do not perceive and therefore do not control, but which are aspects of what is now a collectively controlled variable (your computer functioning). Suffice to say that the variables that she controls which (when controlled well) result in you perceiving that your computer is fixed, and the variables that you control when you perceive that your computer is fixed, are largely not the same variables; I did stipulate that you are not a computer technician.

Within the participating population are differences in what aspects of the collectively controlled variables are perceived and what aspects of the collectively controlled variables are affected by control processes.

Continuing on, her perception of your satisfaction is not your feeling of satisfaction, payment of money is an asymmetrical transaction, and so on. These variables and more are controlled by feedback paths that pass through the functioning of your computer.

Here, it is important to recognize that controlling some distal variable by a feedback path that passes through a collectively controlled variable is a part of, and a form of, participation in the collective control of that variable. To see this, note that disturbance of the collectively controlled variable is resisted when it consequently disturbs your control of the distal variable via the collectively controlled variable.

So start thinking about examples of collective control that don’t involve conflict. Oh, right, you deny that this is possible.

Kent denies this. He probably knows his model of collective control better than you do.

It is evident from your words (such as those quoted above) that you are constraining the word “model” to computer implementations, and that for you “model” is a computer implementation of the PCT model of a behavioral situation. Because there are no other computer simulations of collective control, for you the words “Kent’s model” refer to Kent’s demo programs in the early 1990s.

There’s an important distinction between the PCT model of a behavioral situation and a computer implementation that simulates that situation by replicating data measured in it.

A revision of your statement would say

I think Kent’s simulations of conflict are excellent, but I think he takes the wrong message from them. The message he takes from his simulations is that conflict can produce “social stability” in the form of a variable being kept at a virtual reference level. The message I take from his simulations is that conflict can be very destructive to the individuals involved.

You are saying that Kent arrived at his understanding of collective control by generalizing from his computer demos of conflict. I believe Kent arrived at his understanding of collective control by many years of observing social phenomena with ‘control theory glasses’ and that his computer demos were a first demonstration of principles.

Kent has acknowledged that conflict is a limiting case involving very small populations (two, three, or few). Even within that constraint, his demonstrations of principle in the early 1990s are extremely artificial. Without intervention of higher levels of control conflict does result in ‘stabilization’ of the contested variable at an intermediate level from which in each contestant there persists an error signal and a control output affecting the variable. But higher levels of control are never absent in living control systems. In arm wrestling or a tug of war higher level control of participation in the game overrides other systems (no doubt quit archaic evolutionary inheritances) which on meeting such resistance would ordinarily control alternative variables instead, as illustrated by the video I posted of W at the locked door in Shanghai.

But from that beginning 30 years ago Kent has gone on to develop how PCT can model more complex social arrangements and interactions. Working with him, Martin has shown logical and mathematical considerations that must be taken into account for building computer simulations of more complex examples. However, no one with the necessary skills has been interested in building simulations and demos of such complexity.

Kent has affirmed that he lacks computer skills and the knowledge and training in the design and execution of psychological experiments.

In addition, the requirements for experiment wrt social interactions are very different from the requirements wrt monadic control phenomena such as tracking and pursuit. This fact you have ignored. Instead, you simplify social phenomena to a form that is comfortably within the scope of simulation programming that is most familiar to you. This simplification is evident in your perceiving every example of collective control as a conflict and in your substitution of region (i.e. proximity) in place of esteem in Labov’s findings.

No one will fault you if you are uninterested in tackling this. But if to justify this to yourself you need to deny the existence of social phenomena don’t expect very many people to believe you. Maybe naive extremes of ‘libertarians’ and ‘rational agent’ economists will find it convincing.

RM: …Indeed, one reason I have repeatedly given for not being a fan of Kent’s model is because I can’t think of any examples collective control that involve conflict (other than those I’ve mentioned: arm wrestling, tug of war and war itself). So I can’t think of any interesting examples of collective control phenomena to which his model applies.

Why do we need democracy, parties, elections and elections campaigns? I think we need them and/or they exist because we human beings as members of a society inevitably have different variables we control (with different gains) and different references for similar variables. From these differences follow conflicts: direct conflicts from different references for same variables and more indirect (resource) conflicts from different controlled variables. There could be different ways to solve these conflicts. One possibility is the Hobbesian war of all against all where the strongest participants may oppress or kill their opponents. The Hobbesian story continues that because the results of these wars are insecure and because all control for safety people made a social contract and gave the highest power of decision to the sovereign ruler. Did this end the conflicts? No, I don’t think so, even though it stopped the internal war and stabilized the current situation of oppression (and changed the wars external between the sovereigns). But back to democracy: We use to have elections every few years where we elect our representatives to the sovereign administrative organs like parliaments. Does this end the conflicts? No, I don’t think so. Democratic society is much less oppressive than feudal because in principle everyone can say their word but this does not mean that everyone can as well control their variables. Just because the conflicts remain we need repetitive elections. The war of all against all is now the almost continuous electoral campaign – it is expensive but not as expensive than the war.

So we have continuous conflicts in our society. Can you imagine it? Are our societies instable? Not very, not most of the time. Is it easy to change the virtual reference values of those conflicted variables? The parties try it all the time. Is the situations destructive for us? Do we suffer all the time from the five hundred-fold error? Perhaps we do, but we don’t seem to care much about it. Many feel that it gives them strength – for the next campaigns.

Eetu

To me, the speculation about different input functions, different reference specifications and hierarchical control is irrelevant to whether or not something is collective control. I call it “collective control” whenever more than one control system agent is involved in the controlling.

My “simplification” was just to show that a car being pushed out of the snow by several agents is an example of collective control without conflict to which Kent’s model would apply. But I can’t think of many others. Most of the collective controlling that I see going on – such as the examples described in the Social Control chapter of SCLS --involve no conflict. So I see Kent’s model of collective control as having very limited application.

I think a model that gives a much better idea of how PCT can contribute to our understanding of collective control is Bill’s CROWD model. He built this model to show sociologists how to model collective behavior using PCT. My guess is that this model didn’t catch on because the prevailing zeitgeist in sociology demands a belief in a “collective level” explanation of collective phenomena. Bill’s approach (which, I believe, would have to be considered the PCT approach;-) explains collective phenomena in terms of the controlling done by individuals; there is no extra-individual expanation of the collective phenomena explained by the CROWD demo (forming rings, queues, etc.)

My guess is that Kent’s perfectly good model of conflict caught on as the model of collective control (or the collective control model) because the concept of a “virtual controlled variable” seems like the kind of extra- individual explanation that fits into the sociological zeitgeist (or paradigm).

I have no idea how you came to that conclusion. My point throughout this discussion has been that most examples of collective control I can think of are non-conflictive. All the examples of collective control described in The Study of Living Control Systems (SCLS) are non-conflictive: lifting a couch, flocking, divergence of pronunciation, etc.

I can think of some significant examples of collective control that are conflictive, such as creating legislation. But by and large the collective control I see is mainly cooperative, either intentionally or unintentionally (as discussed in SLCS).

What exactly is it that Kent denies? That it is a model of virtual control? That it is a model of conflict? Or that it is a model of virtual control that results from conflict? I can’t believe Kent would deny what his model clearly shows.

I didn’t know that. What is the distinction?

That’s fine with me. I consider the models (simulations if you like) and PCT to be identical. Modeling is central to PCT. As is testing the models against data. Whether you can do modeling or not, you have to understand the modeling and how those models map to actual behavior in order to be able to do a proper PCT analysis of any kind of behavior: individual behavior, collective behavior, abnormal behavior, etc.

No, I’m saying that Kent’s virtual control model applies to only a small subset of what I would call “collective control”. For example, it doesn’t apply to the flocking of birds or the development of regional dialects.

No more artificial than any of the demonstrations of principle developed by Bill Powers, many of which have been brilliantly re-written as on-line demos by my best old ex-friend Adam Matic.

Kent’s demos were excellent demonstrations of many of the principles that are involved when two control sytstems are trying to control the same perceptual variable. My only complaint about Kent’s modeling is that he hasn’t shown how this model applies to any particular example of collective control. The best way to do this would be to show how the model accounts for data collected from some relevaant example of collective control.

Well, he had enough skill to develop his marvelous spreadsheet models of conflict. I think the appropriate next step for him would be to test this model against data before going on to develop more complex models. He might find that he doesn’t need all the complexity that you imagine is needed. Look at all the different kinds of collective controlling that can be explained by the relatively simple control models that are implemented as the agents in the CROWD program.

I didn’t ignore them becuase I didn’t know they existed. In fact, I don’t think that there are different requirements for social interactions and “monadic” control phenomena. Please tell me what the difference is.

I hope you now know that I don’t perceive every example of collective control as a conflict. Indeed, as I said before, I can’t think of very many that involve conflict. More importantly, I can’t think of any collective control phenomena that involve a conflict that keeps a variable in a virtual reference state. Can you?

And I didn’t substitute region for esteem in Labov’s findings. I picked region because I could assume that people in the same region would be talking with each other more than with people in another region. I could have picked esteem rather than region as the variable that had to do with differences in pronunciaiotn but it doesn’t seem like that has much to do with who you talk to most often and it seemed to me that my best bet for accounting for Labov’s lovely data was a “control for imitation” model.

What is “this” that I am uninterested in tackling?

I don’t deny the existence of social phenomena (see SCLS ch. 7). And I don’t expect – or want – anyone to “believe” me; I want them to understand me. But I don’t expect that either. I know I’m probably just whistling in the wind here but I enjoy doing this and I feel I owe it to Bill Powers; I know I’m whistling his song, but perhaps a little more shrilly.

Yes, these are excellent examples of collective control that involve conflict. And the virtual control model could be a start at understanding these processes. But I think the difference between what happens in producing legislation – the main aim of political activity – and what happen’s in the virtual control model (at least in ostensibly democratic societies) is that the participants are (or should be) willing to take into account the pain (“error signals”) of the other participants as well as their own.

Ideally, successful democatic conflict resolution depends on all members of the collective having an interest in minimizing the overall “hurt” produced by legislation so that the resulting legislation – which is the “virtual reference state” of the variable controlled by the collective – is not simply a result of everyone trying as hard as they can to get what they want (as is currently the case in the virtual control model).

But I think there is a role for the virtual control model to play in the study of producing legislation; it could be a model that shows what would be expected in a “worst case scenario”, where it’s every agent for themselves; capitalism in tooth and claw. The result produced by the model could be compared to what actually happens. And that could be further compared to what would be the “best case” result if everyone cooperated to that each person’s error is simultaneously at a minimum.

Coincidentally, Tim Carey, who understands PCT as well as I do (because he knows that behavior IS control, the essential first step to understanding PCT) has a wonderful blog article that is very relevant to this discussion. I highly recommend it.

Yes, indeed. And I think PCT science – the approach that involves comparing model behavior to data – could contribute to understanding the best way to solve these conflicts. I suspect that any approach to minimizing the overall error in legislative conflict will require that everyone in the collective have a goal somewhat like the one stated by Powers’ and quoted in Tim’s blog article: “The childhood of the human race is far from over. We have a long way to go before most people will understand that what they do for others is just as important to their well-being as what they do for themselves.”

Unfortunately, the childhood of the human race is, indeed, far from over, though. Which is why we still have free-market afficionados who are committed to the idea that what they do for themselves is just as, if not more, important for the well being of others as it is for themselves. That is, they believe that greed is good. And they’ve got the money and power to prove it.

I think one important thing has been forgotten or dismissed in this discussion about collective (or social) control, namely the (source of) disturbance or the “common enemy”. I think that into the models should be added as one more “agent” the source of disturbance which its own gain and reference. Of course it is not a real controller but it can behave as if it were. I mean that the (virtual or not) controlled variable has some (stable or changing) value also when the real controllers are not controlling it and this value is as if the disturbances reference.
Here and there in the literature and discussions is mentioned “overwhelming disturbance”. In those cases the gain of the disturbance is (much) higher than the available gain of the controller.

For example if you control for a stone to be 1,5 meters above the ground but the stone is too heave and you can lift it only a few centimeters, then the disturbance’s (gravity’s) “reference” is 0 meters and its gain is quite overwhelmingly greater that yours. As a consequence your error is high. But now there comes a friend who happens to control for the same stone to be 1 meter above the ground. You and your friend have conflicting references but together you may be able to lift it at least to that 1 one meter and perhaps even some centimeters above it. Now in spite of your conflict you control much better than you would alone and thus without a conflict. In this case the conflict is beneficial for the controllers because the “conflict” with the disturbance can be at least partially solved by the conflictual collective control.

So there is no strict and fixed borderline between collaborative and conflictual collective control but it depends on conditions. Because of our individual differences which Warren mentioned it seems quite improbable that in any collaboration all participants would have exactly same controlled variables and exactly same references for them. Instead there is always also a lesser or greater conflict. But in any case the control tends to create stabilities which in happy cases are more or less beneficial to all or most participants – even if no one were absolutely satisfied.

Eetu

Actually, there are two sources of disturbance that were neither forgotten nor dismissed in the discussion of Kent’s collective control model. One was the active disturbance of the effect of each agent’s actions on the commonly controlled variable; to the extent that the agents have different references for the state of that variable, what each does to move the variable to their preferred reference state pushes it away from the reference state preferred by the others.

The other source of disturbance was passive in the sense that it was not generated by any agent. The fact that this passive disturbance has little or no effect on the commonly controlled variable is what allows us to see that this variable is being maintained in a virtual reference state.

It’s not clear whether you are describing an active or passive disturbance, but it doesn’t matter since both types of disturbance are already part of the model.

In Kent’s model, it’s not the controlled variable that is virtual; it’s the reference state of the controlled variable that is virtual. And when there are no controllers controlling a variable, the value of that variable is never in a reference state of any kind, real, virtual or whatever. It simply varies along with whatever forces (disturbances) are affecting it.

I’ve seen causal models used to explain control phenomena but this is the first time (since reading Aristotle, anyway) that I’ve seen a control model used to explain a causal phenomenon.

Yes, that is demonstrated in Kent’s model when the agents have the same reference for the controlled variable.

True, there is no borderline. But there is a gradient between collaborative and conflictual control and it is very steep. The benefits of having more than one person controlling the same variable are greatest when there is zero difference between the references of the agents involved. As this difference increases these benefits decline very quickly to become huge liabilities with none of the agents getting the commonly controlled variable into their desired state so that the error in each agent increases exponentially.

We’re not talking about modeling physics or environmentally present living control systems here, we’re talking about perceptions constructed within LCSs and projected by them into the environment as though ‘real’. (We all do this all the time, of course, with varying degrees of justification.)

Perception of an “enemy” is very often a collectively controlled perception. Whether or not collectively controlled, it is a perception of a source of past, present, or potential disturbances to one’s control.

In polemics against S-R psychology it’s been important to say that the supposed stimulus (the disturbance) need not even be perceived for the CS to resist its effect on the CV. But of course the disturbance may be perceived, it may be traced to its source, and effective control may be directed there rather than to the CV.

The perception of an “enemy” attributes agency to a source of disturbances. The attributing of agency by the annoyed control system is quite independent of whether the source of disturbance is a ‘real’ control system, a virtual control system, or not a control system at all.

Personally, I’d be unlikely to attribute agency to a big stone, and to perceive that difficulty lifting it was due to its active resistance to my disturbing it. If I were trying to move it out of its embedment in the ground I might say “this stubborn stone just won’t budge!” but this can only be jocular, not a serious attribution of agency, if we think about it and place it within the collectively controlled perceptions that constitute our culture. But we do commonly talk that way. “This doggone ___ won’t cooperate!” And if we are attentive to our own subjective experience it might be difficult to deny there is an impulse to attribute agency, especially adverse or malign agency, contrary to our scientific understandings.

My cat is often enough mistrustful what some inanimate object might do, but how to discern whether some perception of animacy is in play. Researchers have attempted to tease this out. Have they succeeded? For example:

Abdai, Judit; Bence Ferdinandy; Cristina Baño Terencio; Ákos Pogány; & Ádám Miklósi (2017). Perception of animacy in dogs and humans. Biology Letters 13.6. DOI: 10.1098/rsbl.2017.0156

Hofrichter, R.; Siddiqui, H.; Morrisey, M. N.; & Rutherford, M. D. (2021). Early Attention to Animacy: Change-Detection in 11-Month-Olds. Evolutionary Psychology , 19.2. DOI: 10.1177/14747049211028220

Diane Poulin-Dubois, Anouk Lepage, Doreen Ferland (1996). Infants’ concept of animacy. Cognitive Development 11.1:19-36. DOI: 10.1016/S0885-2014(96)90026-X.

Attributions of animacy have a fuller place among the collectively controlled perceptions of other cultures.

The nature of power in the universe is best understood in terms of four basic philosophical assumptions shared by most native Californian groups in their world view… (1) power is sentient and the principal causative agent in the universe; (2) power is distributed differentially … and possessed by anything having "life: or the will “to act”; (3) the universe is in a state of dynamic equilibrium in relation to power; and (4) man is the central figure in an interacting system of power holders. […] Power is sentient and possesses will… potentially extant in all things… Even seemingly inanimate things may possess power. A rock which suddenly moves downhill may thereby demonstrate an ability “to act” and therefore reveal itself to be a power source. […]

Bean, L. J. (1975). Power and its Applications in Native California. The Journal of California Anthropology, 2.1:25-33; repr. in Lowell J. Bean and Thomas C. Blackburn, eds. (1976) Native Californians: A Theoretical Retrospective. Ramona, California: Ballena Press, pp. 407-420 URL: Power and its Applications in Native California on JSTOR

We are? I thought we were talking about PCT.

Huh?

Do I have to go into the archives and find places where Bill wrote “about perceptions constructed within LCSs and projected by them into the environment as though ‘real’” in order for you to accept that we’re talking about PCT?

Of course not. Just show me where "perceptions constructed within LCSs and projected by them into the environment as though ‘real’” are to be found in the PCT model. But it would be nice if you could find where Bill said something like this.

Oh, and it would be helpful if you showed why, when we’re modeling “collective control”, we are not “modeling physics or environmentally present living control systems.”

Hmm, thanks for interesting comments, but I am afraid you both somewhat misinterpreted my intentions. I wanted to stress that in different cases disturbances behave differently – or perhaps I should say that the environment disturbs control differently –regardless of the animacy of the sources of the disturbances. The position of a light stone is easy to control for a single controller but the position of a heavy stone may be impossible to control for a single controller but possible for multiple controllers together. And when they control together it is possible that even if their references for the position were not identical they can reach a lower error level together that they would individually. And similarly very different kind of variables can be more or less heavy or stiff or recalcitrant and hard or even impossible to control individually but possible to control collectively.

In Rick’s demo one controller reach very low error level when controlling alone, but five times higher when controlling in a conflict with another controller. Is it possible to make a model which behaves according to the lines described above and where the disturbance were overwhelming for one controller but a collective of two more or less conflicting controllers could control with lower error level?

I think this is dealt with in PCT by distinguishing between the source of the disturbance, q.d, and its effect on the controlled varilable, h(q.d) (using the notation in Powers, 1978). I’ll use your example to show what this means:

In this example, q.d is the force of gravity, which increases with increasing mass of the stone being lifted. The controlled variable, q.i, is the height of the stone above the earth. The reference state of this variable, q.i* is something like “waist high”. To bring q.i to the state q.i* the controller must produce an upward force output, q.o, greater than the downward force exerted by the force of gravity, h(q.d) on q.i. So the state of q.i (the position of the stone to be lifted) can be written as:

q.i = g(q.o) + h(q.d) (1)

where g() is the feedback function relating output to input. This equation is equivalent to equation (2) in Powers (1978). Assume for now that g() and h() are multipliers of 1.0. So

q.i = q.o + q.d (1a)

For a small stone, the effect of gravity, q.d, on the stones position, q.i, is small so the controller can generate a force, q.o, that is nearly equal to q.d and q.i is easily brought to and held in the reference state, q.i*. But for a large stone q.d is much larger and the controller is unable to generate an output force, q.o, sufficient to lift it – bring it to the reference state, q.i*.

When other controllers help to lift the heavy stone, the controller is now able to lift it because the effect of gravity, h(q.d), on the position of the stone, q.i, has been reduced by the upward force exerted on the stone by the other controller(s). So from the point of view of of the controller who had easily lifted the light stone, the force of gravity on the heavy stone has been reduced.

The other controllers have effectively changed the disturbance function, h() – the function that determines how the source of the disturbance, q.d, affects the controlled variable, q.i – from a multiplier of 1 to a fractional multiplier, say .5. So now the state of the controlled variable is determined by an equation like this:

q.i = q.o + .5* q.d (1b)

I think it is generally the case that the effect of a disturbance on a controlled variable is mitigated like this by another control system, either one that is in the organism controlling q.i or ones that are in other organisms, as in your example.

I think you can show something like this with that “Virtual Control” demo by simply making the references of the two control system equal. In that case you will see that the single control system experiences much more error when it contorls alone than it does when the other system is cooperating rather than conflicting with it.

I think the most important distinction between different types of disturbances that is made by PCT is that between disturbances produced by inanimate (cause-effect) systems and those produced by animate (control) systems. We often mistake inanimate sources of disturbance – like the wind blowing something into your eye – for animate ones. You know you are doing that when you get mad at the inanimate source – the wind in this case – for disturbing you, as though it were an intentional affront. But the difference between inanimate and animate sources of disturbance is important because inanimate sources don’t cause conflict; animate sources do!

Construction of higher level perceptions from lower level perceptions is pervasive in the model. There’s no particular “where it is in the model”.

Subjective experience of intensities, sensations, configurations, transitions, relationships, etc. as ‘real’ phenomena in the environment is not in the model. The model explains subjective experience. No model contains that which it explains.

Sure it is. It’s in the part of the model that shows many lower level perceptual signals entering higher level input functions. I had no problem with this part of what you said.

What you had said is that constructed perceptions are projected into the environment as though ‘real’. This is not a part of the model and it is not something that the model explains. The intensities, sensations, configurations, transitions, relationships, etc. are not explanations of subjective experience. They are hypotheses about the aspects of the environment that organisms control.The model explains the observed fact that organisms control various aspects of their environment. The idea that PCT explains subjective experience completely misses the point of PCT, which is to explain the controlling done by organisms.

Yes, I think we all have a high degree of faith that our perceptions are truthful. You once expressed your faith with this excellent credo:

Rick Marken (2004.07.16.1130)

So that’s the rock (the one that Sam Johnson kicked), but this is the hard place: your only source of knowledge is your perceptions. We understand two senses of the word ‘perception’. Perceptual signals are theoretical entities in the model which have some considerable confirmation as CVs in neuroscience. The experience of perceptions is not theoretical, but it is necessarily subjective.

The work of science is to make the subjective objective, to identify perceptions that we justify calling observed facts. But however factual and real we judge them to be they are still perceptions, subjective experiences.

Hypotheses are subjectively experienced perceptions. To test hypotheses we act as though what is hypothesized is a fact. You have often quoted Bill’s example of the taste of lemonade. We know there are sensors for specific tastes and textures, and the taste of lemonade is a mixture of some of these. Another is the color purple. We know there are cone cells sensitive to green, red, and blue (in a human who is not color blind), and purple is a mixture of stimulation to blue- and red-sensitive cells. To illustrate this, you have created diagrams with variables {v1, v2, …, v3} in the environment, combinations of which correspond to the taste of lemonade, the color purple, etc.

But the variables {v1, v2, …, v3} are controlled variables in the physical sciences. On the authority that we assign to science we assume that these variables are actually present in the environment. But ultimately they are subjective experiences of perceptions by individual scientists.

We perceive that some perceptions are objective, rather than subjective, to the extent that we successfully control them. More so when we agree with one another that we are controlling the same variables. Sciences do this in a disciplined way. Everybody does this informally. “Mmm! Doesn’t this lemonade taste good!” “Look at the chiaroscuro effect here, shading into such a rich, deep purple.”

What makes observed facts objective is the observed fact that the environment is in relevant ways the same for all relevant observers.

Not at all. To say that the model explains subjective experience (which is all that we ultimately have) in no way denies that the model explains other things as well.

The observation of control still has its basis in the subjective experience of the person who projects the hypothesis that their own perception of an aspect of the environment has a reference value to which, when they disturb that variable, it consistently returns as a consequence of what the observed organism does. This is a fascinatingly more complex situation than the color purple or the taste of lemonade, or any other interaction with something that does not control, but it involves the same phenomenon of experiencing a unitary perception (lemonade, purple, control) which theory (a construct of perceptions) tells us depends upon combinations of variables. An important purpose for constructing such theories (the physical basis of the color purple or the taste of lemonade or control) is so that we can justify a claim that these subjective experiences are experiences of realities in the environment. Closer to home, so to speak, these theoretical understandings can help us control better.

A virtual controller is ‘environmentally present’ in the same way. The color purple has a fairly simple relation to CVs controlled by physicists. To observe control requires a much more complex interaction with physical phenomena, projecting the PCT model into the environment on the hypothesis-testing assumption that it is real. Collective control is a yet more demanding hypothesis, but the process of testing it is fundamentally the same. All three cases of scientific explanation are ultimately grounded in individuals’ perceptions—their subjective experiences. In all three cases a scientific model explains those subjective experiences, as well as explaining any perceptual constructs at intermediate levels of CVs (rod and cone cells, wavelengths or photon energetics, levels and interconnections in a PCT model of the behavior, etc.).

You’re just unwilling to look because you deny it’s possible. This 2007 exchange with Richard Kennaway is one place among many. An excerpt:

Bill Powers (2007.02.14.0750 MST)

‘Invention’ or ‘projection’, same thing. We act as if our subjective experiences of the environment are realities in the environment because we have to. Or to turn that around, to be convinced that our perceptions are not veridical would be profoundly disturbing.

To keep this on topic, collective control is an important way of testing and confirming the veridicality of our individual, subjective perceptions. It may be the only way out of solipsism. Johnson had to demonstrate his argumentum ad lapidum to Boswell. Logically, that argument is (informally) fallacious. (See Appeal to the stone in Wikipedia.) That’s why agreement alone is not sufficient. But science rests on replication, in principle if not always in practice, and whether or not something is a replication is a matter for agreement.

Concern about whether or not perceptions are “truthful” is irrelevant to PCT. Actually, it’s more than irrelevant; it’s misleading. It implies that our ability to control depends on having perceptions that are accurate representations of external reality. In fact, our ability to control depends only on having accurate analogs of the perceptual variables we control.

That was in a discussion about whether PCT was solipsistic. It’s not.

I would say that the work of science is to find explanations, preferably in the form of quantitative models, of why we perceive what we perceive. Those models are what we take to be the external reality that is the basis of what we perceive.

The third sentence (and everything after it) appears to be a non-sequiter. It has nothing to do with testing hypotheses about the types of perceptions people control. And I’m quite sure that scientists knew that color perceptions depended on the mixture of light of different wavelengths well before they knew about rods and cones in the retina.

They are simply physical variables. They are not controlled by scientists because they are theoretical entities. Like all other people, scientists control perceptions that are presumed to be functions of those theoretical entities.

Again, those physical variables are purely theoretical. We don’t treat them as actually present in the environment because of the authority we assign to scientists; we treat them as physical variables because they explain the results of experiments so well.

I personally don’t experience the world as a perception; it’s all reality to me. And I don’t have to control things to experience them as objectively real. For example. I can tell that the lamp on my desk is real just by looking at it; same with the books, the desk, etc. Even the after effects of staring at the stupid lamp look objectively real. My experience is of the real world. Intellectually I know that it is all perception but I certainly don’t experience it that way. And I bet everyone else experiences it that way too. Even you.

But the model doesn’t explain subjective experience. Thinking that it does is, as I said, missing the point. The model explains control (purposeful behavior) in terms of control of different types of perceptual variables. What the controller’s subjective experience is while controlling is completely irrelevant to the model. Once I’ve found that you are controlling the distance between cursor and target, for example, I know the perception you are controlling (c - t). I don’t know if you are experiencing this perception in the same way I am but that is irrelevant to the model (and me, for that matter).

This is simply not true There is no “projection” of a hypothesis when testing for controlled variables. It certainly isn’t happening when the computer does the test for the controlled variable in my MindReading demo. Unless you think computer programs “project” perceptions.

I’m sorry. This is just too complicated for me. I really can’t see how this explains your claim that "when we’re modeling “collective control”, we are not “modeling physics or environmentally present living control systems.” Aren’t there people involved in collective control? Aren’t we modeling the behavior of people? Aren’t people living control systems? Aren’t they present in the environment with other people when they are collectively controlling?

What I’m denying is that there is anying in PCT about living control systems projecting their perceptions into the environment. I don’t deny that it’s possible but I’ve never seen any evidence that it happens. But whether it happens or not, I know for sure that projection is not a part of the PCT model.

Bill was talking about a person who was trying to convince herself (and, possibly, others) that there is a reality behind our perceptions. He was not describing projection (or its new synonym, invention) of reality as a property of the PCT model.

I think the only people who need a way out of solipcism are philosophers, and among those, only the ones who want an out from solipsism; post-modernists seem to be quite comfortable with it;-) But solipcism is not a problem for PCT any more than it is for any scientist. Simply assuming that there is a reality behind our perceptions seems to work pretty well in termns of developing successful theories.

PCT can explain why a person (like the one Bill took the role of in his post) might want to invent (or, project) “proofs” that there is a reality behind our perception. But that is the only way that invention (or projection) would fit into PCT. But invention/projection is not part of the PCT model that explains the behavior of a person who does the inventing/projecting.

You search for ways to show that we do not agree, and then assert that I am wrong and you are right. Your persistence in this over many years, with many people, suggests that finding disagreements and then demonstrating that you are right is important to you. Possibly you now blink and say yeah, of course, doesn’t everybody? No, people sometimes make other choices.

To seize upon verbal differences and refute them piecemeal as they occur polarizes apparent contrasts at each point independently. This often has the effect of distorting the relationship of that bit to its context, and sometimes has the effect of changing or even reversing its intended meaning.

An example:

When was the last time you conducted an experiment in physics or chemistry or biophysics or biochemistry? Scientists routinely rely on the scientific consensus in other fields that are epistemically prior to theirs. That is what the words “the authority that we assign to science” mean. (In passing, note that a scientific consensus is a collectively controlled perceptual variable.)

You and I agree that these physical variables are theoretical entities in physics, chemistry, etc. We agree that we treat them as physical variables because they explain the results of experiments. These variables may help explain biophysical experiments into the generation of neural impulses by environmentally sensitive cells in sensory organs. Those are not our PCT experiments. Those are experiments by other scientists. PCT experiments have been consistently at higher levels of perception than perceptual intensities generated by sensors. We refer to intensity perceptions as being at the foundation of the perceptual hierarchy, but that is a gesture in principle, the practice is in prior sciences.

Rather than seeking out and magnifying our differences, I propose that we start by identifying our agreements before proceding to questions with less clear answers. A project of seeking out error in anyone’s understanding of PCT and correcting it is not the only difficulty before us. We know that understanding PCT takes time and persistence and depends upon prior understandings that may be in the way, differently in different people. For this, reiterating and emphasizing what we agree on is important. The difficulties before us in this topic are in questions to which PCT currently does not have clear answers.

Here are some things that I think we agree on:

  1. That it’s all perception, we have no privileged access to what’s going on other than our perceptions.
  2. That higher-level perceptions are functions of lower-level perceptions.
  3. That at the lowest level biophysical stimulations of sensors are functions of we-know-not-what in the environment.
  4. That various sciences have models of what is in the environment, which help to explain how our sensors and nervous systems construct our perceptions.
  5. That those models are themselves high-level perceptions.
  6. That the physical variables {v1 … vn} to which we refer to account for intensity perceptions at the lowest level are theoretical entities within explanatory models controlled in the practice of the physical sciences.

I will be surprised if you disagree with any of this. If you find some words to carp about then we can drill down to the intentions.

Now let’s consider some matters that are less clear.

Perceptual signals are theoretical variables in the model. Our actual perceptions are subjective experiences. Subjectively experienced perceptions are the only reality we know. As Bill put it in his post “PCT is about experience” (the origin of the ‘taste of lemonade’ example):

Experiences are not in the PCT model. What is in the model are quantitative relationships. Quantitative relationships are theoretical (e.g. the loop reduces error asymptotically to within a tolerance range corresponding to loop gain). Specific quantities are not in the model either. Specific quantities are experimental data and data displayed by working simulations of experimental results. None of this escapes the fundamental condition that it’s all perception.

We talk as if perceptual signals in neurons were our perceptions, as if theoretical entities in the PCT model were our experiences which are being modeled, but we know better. Or we should. This does lead to unfortunate confusions for students, as when people talk of experiencing or feeling error, when what they usually mean is (according to Bill’s model of emotion) that they experience perceptions of somatic conditions that result from unreduced error, which may feel distressing, and they experience higher-level perceptions which we call emotions, which include the somatic states and feelings of distress among their inputs.

Perceptual signals and all the other objects and relations in the PCT model are also in the universe of subjectively experienced perceptions. It’s all perception, and the only perceptions we know are subjective experiences, not perceptual signals. The PCT model tells us they are perceptual signals. Experiment tends to confirm the PCT model. Experimental confirmation in the epistemically prior sciences of neurology and neurobiology still lags.

You object to the phrase ‘projected into the environment’. The 1992 quote above is fairly early in the ‘mirror world’ discussions of how the experienced world is the perceptions constructed in the perceptual hierarchy, but experienced as intensities, edges, objects, relations, etc. present in the environment. That is what the phrase ‘projected into the environment’ meant then and that is what it means here. PCT is about experience. It explains experience.

‘Projecting’ is experiencing internally constructed perceptions as though they are present in the environment. The computer program has no experiences. Living control systems do. PCT currently has no explanation for the difference. ‘Projecting’ experienced perceptions as though perceived environmental phenomena is not necessary for control (viz. various kinds of automata).

Yes, I agree. My subjectively experienced perceptions are the only reality I know. Among my subjectively experienced perceptions are system concepts, principles, and procedures of various kinds that make a distinction between more objective (the lamp) and more subjective perceptions (the lingering perceptions of light intensity and form after looking away). Importantly here, theoretical and experimental work in science are ways of controlling subjective perceptions on the basis of which we perceive more confidence that they are objective. That kind of difference is the distinction I was making.

That last sentence seems backward. Our ability to control tells us that the perceptual variables that we control are sufficiently analogous to the environmental sources of those perceptual variables, whatever those environmental sources may be. But we experience the taste of lemonade, the color purple, and someone saying something wrong about PCT, as though these were actually present entities in the environment.

(We do act in ways that appear to make environmental realities more closely analogous to perceptual variables that we control, but that’s the negentropic nature of control. Not the point under consideration here.)

Your mentioning that context does not disqualify it from this discussion. It is true in any context, isn’t it?

No one is contesting the fundamentals of PCT. This is a discussion of questions to which PCT currently does not have clear answers. The model is necessary for explaining quantified data of behavior which are understood to be numerical analogs of perceptions. But people also look to PCT for understanding of what they experience. And not clinicians and their patients alone. The relevance of PCT to subjective experience is crucial to the spread and adoption of PCT. In that process of spread and adoption, to minimize and correct the distortions and misconceptions to which learners of PCT are so abundantly prone we must address those learners where their interests lie, and that means we need a coherent account of the relation of the quantitative theoretical objects and relations in the model to people’s subjective experience. The relation of perception as signal to perception as experience is right at the heart of the famous old mind-body problem. Some folks pretend that it’s a pseudo-problem, that it doesn’t exist because mind and consciousness are epiphenomena, nothing but matter all the way down (or up). Skinner headed that way. Thomas Huxley and William James fenced over this. Frankly I doubt that anyone really believes it. Like the determinists they just say dourly that’s what logic compels them to believe and are unable to reevaluate their assumptions.

You speak of facts. Facts are perceptions to which we give credence for good and sound reasons (and those reasons are also perceptions). Here’s a voice from 2014:

I agree that perceptions (including facts) can be of different levels of complexity. I indicated several levels of complexity or abstraction. [1] The color purple or the taste of lemonade are perceptions at a low level of complexity, and [2] the PCT model is a perception at a high level of complexity. (In PCT we can refer to their complexity e.g. according to their level in the hierarchy, but elsewhere this distinction is expressed by words like concrete or material vs. abstract, conceptual, theoretical.) In both cases these are perceptions closed by loops passing through one control hierarchy.

[3] The third degree of complexity is when control loops through several control hierarchies are closed through intersecting environmental feedback paths, so that they non-identically influence physical properties of the same aspects of the environment, of which they control their non-identical perceptions. Of this kind of complexity I gave as an example a virtual controller, a theoretical entity in modeling collective control.

[quote=“rsmarken, post:49, topic:16019”]

Yes. But when you model more than one individual interacting in a common environment, in addition to each potentially influencing variables that others are controlling (purposefully or not), each may control perceptions of some or all of the others, including perceptions (mistaken or not) of others’ purposes. One individual may have responsibility for actively controlling a variable. One may delegate responsibility for control of a variable to another. One may communicate to another that it is their responsibility or their turn or would you please. Such communication may be nonverbal, e.g. ostensive incapacity or unwillingness or non-perception. Importantly, one who delegates control to another has not necessarily stopped controlling the delegated CV. Possible situations among interacting control hierarchies can be quite complex.

Much too broad an interpretation. What I said, a bit more fully, was:

In part mentioning physics was nod to your objection to Eetu’s talk of lifting a heavy stone, but more immediately it led into discussion of a ‘virtual controller’. The theoretical entity ‘virtual controller’ does not exist in physical terms, but it is a useful concept for grappling with the complexities of collective control. Just testing and confirming that a variable is consistently restored to a reference value cannot distinguish between individual control and collective control. Determining where the control outputs come from is more complicated for collective control because the participating control hierarchies may exert control at different times. Verifying that each can perceive the CV, can exert outputs that affect the state of the CV, when doing so is in fact perceiving the CV (i.e. control stops when perceptual input is blocked), etc.—all those steps of the Test in addition to disturbing the CV are more difficult to carry out in an interactive collective control situation.

The part of the PCT model in which we model physics is the environmental feedback function. (We don’t do a whole lot with the psychophysics of peripheral, environment-facing input and output functions, and next to nothing with biophysics, neurochemistry, and neurophysiology within the hierarchy, but they also ground us in physical sciences.) The PCT model has been applied mostly to one individual control hierarchy at a time. In modeling collective control, the concept of a virtual controller models the observed fact that a collectively controlled variable is controlled at a reference level as though one control system were controlling it. Just as the PCT model is abstract, a virtual controller is an abstract part of the PCT model when we consider collective control. In principle, it is possible to resolve the inputs and outputs of the virtual controller to the inputs and outputs of the individual autonomous control hierarchies participating in that collective control, but in actual social situations it may often not be practicable. There are just too many potentially asynchronous variables for the experimenter to bring under control in an experimental design, or for the observer to bring under control in a program of naturalistic investigation.

That’s fine. Nothing compels you to have anything to say about it.