Stability and Control (was Re: TCV and Collective Control ...)

[From Rick Marken (2016.12.09.1750)]

···

Erling Jorgensen (2016.12.08 1150 EST)

[EJ] Rick, I appreciate the archival search you did to find some of Bill Powers’ thoughts on this issue. For one thing, it helps me relax about designating the CV term (Controlled Variable) the Observer’s version of the perception seemingly being controlled. As Bill states:

Bill Powers (961224.1145 MST)

>>BP: Remember that as far as the observer is concerned, what is controlled i****s ONLY the CV. The idea that this CV is represented by a perceptual signal inside the other system is theoretical.

RM: I think it’s more correct to say that the Observer comes up with a hypothesis about what the other person’s CV is. Otherwise, the rest of your description of the TCV is spot on.

RM: All control involves stabilization of variables. In control engineering “stability” is the ratio of a measure of the actual variance of a controlled variable relative to its expected variance if if were not controlled (what we call the “Stability factor” in PCT research). When this ratio is very small – when the actual is much smaller than the expected variance – control is good. As you note, what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good.

RM: So saying that the variable in conflict is stabilized is the same as saying that it is controlled. Of course, what is different about the conflict situation is that the controlled variable is being controlled relative to a virtual rather than the actual reference levels of the systems involved in the conflict. But we know this from a proper analysis and model of the conflict situation. If all we could observe is the behavior of the variable in conflict we would conclude that it is, indeed, a controlled variable, being maintained in a reference state, protected from disturbance.

RM: There is no question that the stability of environmental variables can affect the feedback connection between outputs and controlled inputs. Just ask someone climbing El Capitan; unstable rocks can ruin your day. But here the meaning of “stability” is different than its meaning in control engineering. In this case “stability” refers to the denominator of the “stability factor”; the expected variance of a variable assuming only physics were involved; no control system involved. The surface of El Capitan varies in this kind of stability due to differences in physical forces acting on the rock surface.

RM: So the same word, “stability” refers to two different things. “Stability” can refer to control or it can refer to physical immovability. I think the way to solve this ambiguity is to use the term “control” when we are talking about stability as control and “stability” when we are talking about stability as physical immovability. In Kent’s paper, the word “stability” is used to mean “control”; houses and computers are not just a stable result of human collective action; they are controlled results. El Capitan is a stable result of tectonic forces; it’s tough to move because these forces produced a massive rock.

Best

Rick


Richard S. Marken

“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.” – William T. Powers

[EJ] This says that from the Observer’s point of view, the only thing they have to work with is the CV. In the Test for the Controlled Variable, they build up a hypothesis about the other person’s relation to the CV

EJ: – attempting to control it or not – based on what happens following the Observer’s disturbance of that variable. If the CV moves as expected, then try something else, because that’s not the variable the other is controlling for. But if the CV does not move as much as expected from their own disturbance, that is a significant finding, seemingly because something or someone is keeping that result from happening. When variables are stabilized in that way, we suspect “control”, with a very useful theoretical model for what might be going on. And indeed, you and Bill developed the notion of a “Stability Factor”, to quantify the degree of that hypothesized control.

[EJ] However, I disagree with an earlier point you (and Bill) make. The relevant context is the two sentences prior to Bill’s quote above:

Bruce Abbott (961224.1310 EST)

**>>>**BA: Rick’s response was to deny that the distinction Martin was making between “stabilized” and “controlled” was useful. CV, he said, either is controlled or is not controlled.

**>>**BP: I would tend to agree with Rick, because of my definition of control given above.

[EJ] I have two reasons for considering the notion of “stabilized-but-not-controlled” to be useful. The first reason derives from Kent McClelland’s modeling of conflictive control situations. When two living control systems are trying to keep the same perceptual variable in two different reference states, the result is often a “virtual reference level” somewhere in between the two preferred states, roughly proportional to the relative contributions of each party’s output gain. While neither party achieves satisfactory “control” of the variable, its value is definitely “stabilized” somewhere in the middle, with each party pulling as hard as they can. In fact, this is how Bill Powers used to talk about such situations – when the output of each party is maxed out in this way, they have lost effective control, and the variable may well drift according to whatever other disturbances are in play.

[EJ] My other reason for considering “stabilized” a useful concept that does not simply overlap with “control” was given in my recent post (Erling Jorgensen (2016.12.01 1430 EST). There I argued that a stabilized niche can be an important part of the Environmental Feedback Function for controlling other variables. Here, the term stability refers to properties of a given control loop, not the values generated by the loop itself. As I said in more detail there, “Stabilized properties, for more effective control.”

All the best,

Erling

[From Erling Jorgensen (2016.12.13 1155 EST)]

Rick Marken (2016.12.09.1750)]

RM: In control engineering “stability” is the ratio of a measure of the actual variance of a controlled variable relative to its expected variance if it were not controlled (what we call the “Stability factor” in PCT research). When this ratio is very small – when the actual is much smaller than the expected variance – control is good.

[EJ] Thanks for the clarification of how the Stability Factor is calculated. Yes, this works very well when it is an assessment of an individual system’s degree of control.

RM: …what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good.

RM: So saying that the variable in conflict is stabilized is the same as saying that it is controlled.

[EJ] But here you are begging the question being raised, by simply defining small stability as good control. The situation being discussed is when there are competing or conflicting control systems. In a tug-of-war with each team pulling as hard as they can, neither team has yet “controlled” the flag by bringing it to their preferred position, if the flag is still hovering (“stabilized”) in the middle. Bill Powers’ himself said – (I love Rupert’s affectionate designation, the Big Powowski!) – that control systems in intractable conflict have effectively lost control.

[EJ] With apologies, I’ll raise a recent political example. The emerging decisions of the incoming Trump administration are very far from my control §references, & that of 65.7 millions of others, even though they are stabilized much closer to that of 62.9 million other voters. The argument being raised in this CSGnet discussion is whether “stabilized” is a more accurate way to talk about those “virtual reference level” compromised outcomes between conflicting control systems. In those situations, the meaning of the Stability Factor is being confounded – it is not a good measure of individual control, because of the confound of other (conflicting) control systems.

[EJ] Kent’s key argument in his recent post [Kent McClelland (2016.12.09.1400)], with which I agree, is that “we need to develop a clear analytic vocabulary for talking about the collectively controlled stabilities in our common social environments.” I understand Bill Powers’ reluctance, and I share it, for not going beyond the bare minimum of conceptual categories in PCT. That is why I am always looking for how much mileage we can get out of the classic PCT concepts – e.g., associating conflict as mutual “Disturbance”, noticing stable niches emerging in the “Environmental Feedback Function”, correlating various emotions with the rate of change in the “Error” term, etc.

[EJ] However, there are “cultural stabilities” that develop, like well-worn pathways, which are themselves phenomena worthy of investigation (“Phenomena Phirst!”, as the rallying cry goes). Like Kent, I’m not sure what language is best for construing them. Various candidates have been proposed:

“Atenfel” (aesthetically, not getting my vote).

“CEV / Complex Environmental Variable” (perhaps subject to the representational misunderstanding).

“Stabilized niche” / “Well-worn path” / “Stable platform for controlling other perceptions”.

“Socially produced invariance”.

“Boss realities” that discipline reorganization.

The “Winter leaf effect” of control regions semi-protected from further disturbance.

[EJ] To speak in these terms does not keep them from also being Categories / Principles / System Concepts controlled by individuals in their personal hierarchies of control. It just allows asking broader questions about the properties of control systems as they interact.

All the best,

Erling

[Martin Taylor 2016.12.13.14.32]

[From Erling Jorgensen (2016.12.13 1155 EST)]
>Rick Marken (2016.12.09.1750)]
>RM: In control engineering "stability" is the ratio of a measure of the actual variance of a controlled variable relative to its expected variance if it were not controlled (what we call the "Stability factor" in PCT research). When this ratio is very small -- when the actual is much smaller than the expected variance -- control is good.
[EJ] Thanks for the clarification of how the Stability Factor is calculated. Yes, this works very well when it is an assessment of an individual system's degree of control.

I don't like the term "Stability Factor" because a lot of uncontrolled things are stable. Rick mentioned El Capitan. I might add a weight hung by a spring or a ball in a bowl. The term I prefer is "Control Ratio" for the ratio Rick describes. That's teh wording that used to be standard on CSGnet.

>RM: ...what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good.
>RM: So saying that the variable in conflict is stabilized is the same as saying that it is controlled.
[EJ] But here you are begging the question being raised, by simply defining small stability as good control. The situation being discussed is when there are competing or conflicting control systems. In a tug-of-war with each team pulling as hard as they can, neither team has yet "controlled" the flag by bringing it to their preferred position, if the flag is still hovering ("stabilized") in the middle. Bill Powers' himself said -- (I love Rupert's affectionate designation, the Big Powowski!) -- that control systems in intractable conflict have effectively _lost_ control.

The individual controllers may have lost control, but that does not mean that the position of the "flag" must be uncontrolled. In a tug-of-war, it actually is uncontrolled because both teams are pulling as hard as they can, but in other analogous situation in which they aren't, the "flag" appears to be controlled by a system with a gain that is the sum of the two gains, and a reference value given by the ratio of the gains between the two individual reference values. Unless the tester can actually see the actors, there's no way that the TCV could tell the difference between what I call a "Giant Virtual Controller" and a real physical controller.

[EJ] With apologies, I'll raise a recent political example. The emerging decisions of the incoming Trump administration are very far from my control (p)references, & that of 65.7 millions of others, even though they are stabilized much closer to that of 62.9 million other voters. The argument being raised in this CSGnet discussion is whether "stabilized" is a more accurate way to talk about those "virtual reference level" compromised outcomes between conflicting control systems. In those situations, the meaning of the Stability Factor is being confounded -- it is _not_ a good measure of individual control, because of the confound of other (conflicting) control systems.

Nor is your example an example of two opposed controllers. The situation is a one-shot action to affect the value of a relationship: "Electoral Votes for A minus Electoral Votes for B". There is no feedback loop, no corrective action possibility. The individual actors may be controlling for the result to match their reference values, but they also get only one shot, just as would be the case if a hunter only got one shot at a deer. He has a reference to perceive himself eating venison, but if he misses and the deer runs away, he won't eat. He is controlling for seeing the deer to be not alive, whereas the deer is controlling for perceiving itself to be alive. There's no intermediate condition and no opportunity for the condition to be changed after the trigger is pulled.

[EJ] Kent's key argument in his recent post [Kent McClelland (2016.12.09.1400)], with which I agree, is that "we need to develop a clear analytic vocabulary for talking about the collectively controlled stabilities in our common social environments."

I agree, and I would argue that we need terms that deal not just with collectively controlled stabilities, but with all the socially interactive effects of control, not to mix them up with related concepts in individual control. To use the same words in different domains is just asking for misunderstanding. If the same words properly apply in both domains, then by all means let's use them, but much of the time they don't.

I understand Bill Powers' reluctance, and I share it, for not going beyond the bare minimum of conceptual categories in PCT. That is why I am always looking for how much mileage we can get out of the classic PCT concepts -- e.g., associating conflict as mutual "Disturbance", noticing stable niches emerging in the "Environmental Feedback Function", correlating various emotions with the rate of change in the "Error" term, etc.
[EJ] However, there are "cultural stabilities" that develop, like well-worn pathways, which are themselves phenomena worthy of investigation ("Phenomena Phirst!", as the rallying cry goes). Like Kent, I'm not sure what language is best for construing them. Various candidates have been proposed:
"Atenfel" (aesthetically, not getting my vote).

Nobody has proposed an aesthetically better word. As I said when this last came up, It's the best we could come up with after rejecting several other candidates. It may be the result of an environmental stability or a stability created by individual or collective control, but an atenfel is just a segment of the environmental feedback path between the output of an ECU (Elementary Control Unit -- another term) and the perceptual input of that same ECU. It's in a conceptually different domain, just as a wire is in a conceptually different domain from a voltage.

"CEV / Complex Environmental Variable" (perhaps subject to the representational misunderstanding).

A CEV is not a cultural stability. It's just a variable that is a function of "simpler" variables. When a similar concept refers to a cultural stability (i.e. a collectively controlled value of an environmental function) I call it a CCEV (Collective Complex Environmental Variable, to distinguish the two concepts. A CCEV corresponds to a virtual perception controlled by a Giant Virtual Controller to some reference value that may not be held by any of the individual controllers whose controlling results in the observed effects. Have a look at my CSG93 talk, which Rick said clarified the matter for him in respect of Language as a CCEV, when it had been unclear in our exchanges before the meeting on the precursor list to CSGnet. <http://www.mmtaylor.net/PCT/Movie/TaylorCSG1993.mp4> The relevant part starts 19 minutes 15 seconds into the movie.

"Stabilized niche" / "Well-worn path" / "Stable platform for controlling other perceptions".
"Socially produced invariance".
"Boss realities" that discipline reorganization.
The "Winter leaf effect" of control regions semi-protected from further disturbance.
[EJ] To speak in these terms does not keep them from also being Categories / Principles / System Concepts controlled by individuals in their personal hierarchies of control. It just allows asking broader questions about the properties of control systems as they interact.

True, but that's a Analyst's viewpoint, not the viewpoint of any individual controller or observer.

Martin

···

All the best,
Erling

[From Erling Jorgensen (2016.12.14 0945 EST)]

[Martin Taylor 2016.12.13.14.32]

Erling Jorgensen (2016.12.13 1155 EST)]

[MT] I don’t like the term “Stability Factor” because a lot of uncontrolled
things are stable. Rick mentioned El Capitan. I might add a weight hung
by a spring or a ball in a bowl. The term I prefer is “Control Ratio”
for the ratio Rick describes. That’s teh wording that used to be
standard on CSGnet.

[EJ] I don’t recall the use of that term, “Control Ratio,” but I like it. Perhaps it was used in the earlier days of CSGnet, before I had joined the listserv. I can also understand a judgment call about how “Stability Factor” might be intriguing to a non-PCT readership, without eliciting [sorry about the behavioral framing here] the knee-jerk reaction against the notion of “control.”

[EJ] In a tug-of-war with each team pulling as hard as they can, neither
team has yet “controlled” the flag by bringing it to their preferred
position, if the flag is still hovering (“stabilized”) in the middle.
Bill Powers’ himself said – (I love Rupert’s affectionate
designation, the Big Powowski!) – that control systems in intractable
conflict have effectively lost control.

[MT] The individual controllers may have lost control, but that does not mean
that the position of the “flag” must be uncontrolled. In a tug-of-war,
it actually is uncontrolled because both teams are pulling as hard as
they can, but in other analogous situation in which they aren’t, the
“flag” appears to be controlled by a system with a gain that is the sum
of the two gains, and a reference value given by the ratio of the gains
between the two individual reference values. […the quotation continues further below…]

[EJ] I believe that is the issue being discussed here: I.e., whether “control” is the best word for those interactive situations where there is the appearance of control, but the relevant perception is not sufficiently close to either party’s actual reference. I’m arguing that “stabilized” is better, to emphasize that the outcome may derive from competing systems whose actual “control” is pretty poor, where control refers to tracking their own respective reference level. This doesn’t mean that we can’t still use the whole PCT armamentarium to analyze the situation. (Notice, as well, that “stabilized” as an verb acting in the situation seems better than “stability” as a property of the system.)

[EJ] Perhaps we need to resurrect the old language of “controlling for” a given value of a perception. Maybe we should consider a term such as “agent based virtual control,” similar to the work with Agent-Based Computational Economics that Charlotte Bruun has done.

[EJ] Returning to Martin’s interrupted quotation above…

[MT] Unless the tester can
actually see the actors, there’s no way that the TCV could tell the
difference between what I call a “Giant Virtual Controller” and a real
physical controller.

[EJ] I wonder about this. I’m thinking spatially, & considering the simplified case of just two competing control systems pulling in opposite directions. And I guess it may be important that their respective outputs have not yet maxed out. I’m picturing it as two adjoining circles, representing the two control systems, where the radius represents each one’s output pulling on a common variable, which is stabilized on a tangent point between the two circles.

[EJ] There are various directions an additional Disturbance could be applied in the Test for the Controlled Variable. If the TCV disturbance is strictly orthogonal to both systems (i.e., pulling sideways out the joint tangent line), the output of both systems would be predicted to increase, because the variable gets further away from each of their preferred reference for it. But what if the TCV disturbance is applied along the radius of one of the circles, or in some other manner. Wouldn’t there be a difference in their respective outputs, & could that differential help to distinguish between a single controller being involved versus a virtual controller with more than one control system contributing to the outcome?

[MT] Nor is your example an example of two opposed controllers. The situation
is a one-shot action to affect the value of a relationship: “Electoral
Votes for A minus Electoral Votes for B”. There is no feedback loop, no
corrective action possibility. The individual actors may be controlling
for the result to match their reference values, but they also get only
one shot, …

[EJ] Not to spend too much time on my political analogy, but I view voting as just one output among others towards similar references, including such things as “supporting” a given candidate before the election, & “lobbying” for preferred policies afterwards.

[EJ] …Various
candidates have been proposed:

[EJ] “Atenfel” (aesthetically, not getting my vote).

[MT] Nobody has proposed an aesthetically better word. As I said when this
last came up, It’s the best we could come up with after rejecting
several other candidates.

[EJ] For me, it’s not just aesthetics, but rhetorically, too. While I haven’t read the LCS-IV draft contributions where I believe the full argument is laid out, I’ve tried to notice my own reactions when you or Kent have used that term “atenfel.” I always notice myself slowing down my absorption of the train of thought, while I tried to figure out ‘What were those words again that the letters represent?’ In that sense, “atenfel” gets in its own way. It doesn’t easily build on an evocative or associative context, to get me part of the way there, the way the word “cybernetics” could possibly evoke “steersmanship.” So, for now, I notice myself inserting a term such as “feedback path” when I run into the word “atenfel.”

[MT] It may be the result of an environmental
stability or a stability created by individual or collective control,
but an atenfel is just a segment of the environmental feedback path
between the output of an ECU (Elementary Control Unit – another term)
and the perceptual input of that same ECU.

[EJ] So, how about the term “EFF-segment”? Environmental Feedback Function is already in the vernacular, even though it is not emphasized as much as the other functions in the control loop. I think ‘segment’ is the additional point you’ve now clarified for me, in how you use the concept.

[MT] It’s in a conceptually
different domain, just as a wire is in a conceptually different domain
from a voltage.

[EJ] Yes, this is the distinction I was trying to make by distinguishing the “properties” of the loop from the “values” that travel along the wires.

[EJ] “CEV / Complex Environmental Variable” (perhaps subject to the
representational misunderstanding).

[MT] A CEV is not a cultural stability. It’s just a variable that is a
function of “simpler” variables. When a similar concept refers to a
cultural stability (i.e. a collectively controlled value of an
environmental function) I call it a CCEV (Collective Complex
Environmental Variable, to distinguish the two concepts.

[EJ] I appreciate hearing how you use & define these concepts. I don’t believe there is yet consensual agreement among the PCT/CSG community on the use of these terms.

[MT] A CCEV corresponds to a virtual perception controlled by a Giant Virtual
Controller to some reference value that may not be held by any of the
individual controllers whose controlling results in the observed
effects. Have a look at my CSG93 talk, which Rick said clarified the
matter for him in respect of Language as a CCEV, when it had been
unclear in our exchanges before the meeting on the precursor list to
CSGnet.

[EJ] I believe I was there to hear your talk, because that is when I first met you. I was certainly struck by the Layered Protocols conceptualization you had been working on. And, yes, I do see Language as one of these collectively controlled stabilities that enable or constrain all sorts of other forms of control. I will try to look again at your presentation.

[EJ] Perhaps the key language we are seeking is in that word “virtual.” Speaking of a “virtual reference level” for a “virtual perception” in a “Giant Virtual Controller” at least makes clear that the resulting stability may or may not line up with any individual’s preferred reference level in what they are attempting to control.

[EJ] I believe there is pretty good convergence between us on what we are trying to talk about here. Just some disagreement still on the best terms to use. Thanks for the link below.

[MT] <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.mmtaylor.net_PCT_Movie_TaylorCSG1993.mp4&d=DgICAg&c=YJISzXwipuhs1sOHhq_MVsCfWsQ3M4nmaXNHj0MKUZI&r=DECNPmzi23eNTxgLRzMfqnZp9q9U-RE7SpMagl9-sjQ&m=34flWoW71pY3wKP1pSlu6p3G4rdHmaHQmMH0BV8eZFM&s=TjqEn8t43SNLca37C9KGD9glLxIqr6uSd7oWBsg-Avo&e= > The relevant part starts 19 minutes 15 seconds into the movie.

[EJ] “Stabilized niche” / “Well-worn path” / “Stable platform for
controlling other perceptions”.
“Socially produced invariance”.
“Boss realities” that discipline reorganization.
The “Winter leaf effect” of control regions semi-protected
from further disturbance.

[EJ] To speak in these terms does not keep them from also being
Categories / Principles / System Concepts controlled by individuals in
their personal hierarchies of control. It just allows asking broader
questions about the properties of control systems as they interact.

[MT] True, but that’s a Analyst’s viewpoint, not the viewpoint of any
individual controller or observer.

[EJ] Um, yeah… I think that’s what we’re doing here – trying to work out a clear & useful “analytic vocabulary for talking about the collectively controlled stabilities,” to use Kent’s phrase.

[EJ] To the above list of associated concepts, I think we’ve now added the following, (admittedly, this is a clustering of ideas that may not all have the same referent):

“Controlling for”

“Agent based virtual control”

“Giant Virtual Controller”

“Feedback path”

“EFF-segment”

[EJ] I see this whole discussion as an example of collectively controlled convergence as to perceptions, (well, hopefully convergence), in addition to an analytically useful conversation.

All the best,
Erling

[From Rick Marken (2016.12.14.1030)]

···

Erling Jorgensen (2016.12.13 1155 EST)

RM: …what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good.

RM: So saying that the variable in conflict is stabilized is the same as saying that it is controlled.

RM: No, control means stabilized in the face of disturbance that should produce instability. That’s why the stability factor is a measure of observed to expected stability. Expected stability is a measure of the variability of a variable that would be expected on purely causal principles; observed stability is a measure of the actual variability of the variable. If the observed variability of the variable is much smaller that the expected variability it is evidence that the variable is being protected from the effects of disturbances – ie. that it is being controlled.

RM: What Kent showed is that under certain circumstances, a variable that is in conflict can appear to be controlled. Those “certain circumstances” exist when the systems in conflict are not producing maximum output. But in real conflicts, both systems quickly go to maximum output. When this happens, the variable no longer acts as though it’s controlled; the variable is no longer protected from disturbances. This is why powers said that when system are in conflict both lose control. It’s because disturbances to the variable both are controlling (relative to different references) are completely effective. You can see (and experience) this loss of control in my “Cost of Conflict” demo (http://www.mindreadings.com/ControlDemo/Conflict.html).

RM: So the typical result of conflict is neither stability nor control. There will be stability, in the sense of low variability of the variable in conflict, if there are no disturbances acting on the variable. But if there are disturbances acting on the controlled variable then there will be no stability (the variable in conflict will vary in proportion to the variable effects of the disturbance) and there will clearly be no control (the ratio of observed to expected variance of the variable in conflict will be 1.0, indicating that the variable is not protected from disturbance at all.

RM: So what Kent discovered is that there is a situation where a variable in conflict can be controlled, in the sense that it can be maintained in a reference state, protected from disturbance. This happens only when the systems in conflict are not operating a maximum output, a situation that is surely very rare (I can’t think of any but perhaps you can). A variable can also be stabilized by conflict, but again only in situations where there are no disturbances acting on the variable in conflict. This may be a more common situation and it may be what Kent was talking about when he talks about conflict resulting in stability. Again, I can’t think of any socially significant examples of such situations; the only example I can think of is a tug of war with teams of nearly equal strength. But as the tug of war situation suggests, this stability is always on the edge of instability.

[EJ] With apologies, I’ll raise a recent political example. The emerging decisions of the incoming Trump administration are very far from my control §references, & that of 65.7 millions of others, even though they are stabilized much closer to that of 62.9 million other voters. The argument being raised in this CSGnet discussion is whether “stabilized” is a more accurate way to talk about those “virtual reference level” compromised outcomes between conflicting control systems. In those situations, the meaning of the Stability Factor is being confounded – it is not a good measure of individual control, because of the confound of other (conflicting) control systems.

RM: I don’t see what variable is being stabilized in your political example. So I’ll just say that there is no “confounding” of the stability factor that I can see. Whether a variable is being controlled by 1 or 100 people is irrelevant to the computation of the stability factor. All you are doing is measuring the observed variability of a variable and comparing it to the expected variability. The calculation of the expected variability does require that you know how many people are acting on the variable and what the expected effect of those actions combined with the effect of disturbances to the variable are expected to have if the variable were not under control. But I think this can be very easily done, to quote a phase (there, two Dylan lines in one sentence!!)

RM: I still think what’s really needed are observations – preferably quantitative – of the social phenomena that you want to explain. I think the main problem (for me, anyway) with all this discussion of “collective control” is that the theory has driven what is observed rather than vice versa. And the social phenomenon that has been pushed as the one explained by collective control is stability. Stability is a measure of the variability of of a variable; the lower the variability of the variable, the more stable it is. So is collective control just trying to account for the stability of social variables? Or is it trying to account for the fact that these variables are more stable than would be expected on causal grounds – that is, it is trying to account for the fact hat these variables are controlled?

RM: What are the social phenomena that you are trying to account for with control theory?

Best

Rick

[EJ] Kent’s key argument in his recent post [Kent McClelland (2016.12.09.1400)], with which I agree, is that “we need to develop a clear analytic vocabulary for talking about the collectively controlled stabilities in our common social environments.” I understand Bill Powers’ reluctance, and I share it, for not going beyond the bare minimum of conceptual categories in PCT. That is why I am always looking for how much mileage we can get out of the classic PCT concepts – e.g., associating conflict as mutual “Disturbance”, noticing stable niches emerging in the “Environmental Feedback Function”, correlating various emotions with the rate of change in the “Error” term, etc.

[EJ] However, there are “cultural stabilities” that develop, like well-worn pathways, which are themselves phenomena worthy of investigation (“Phenomena Phirst!”, as the rallying cry goes). Like Kent, I’m not sure what language is best for construing them. Various candidates have been proposed:

“Atenfel” (aesthetically, not getting my vote).

“CEV / Complex Environmental Variable” (perhaps subject to the representational misunderstanding).

“Stabilized niche” / “Well-worn path” / “Stable platform for controlling other perceptions”.

“Socially produced invariance”.

“Boss realities” that discipline reorganization.

The “Winter leaf effect” of control regions semi-protected from further disturbance.

[EJ] To speak in these terms does not keep them from also being Categories / Principles / System Concepts controlled by individuals in their personal hierarchies of control. It just allows asking broader questions about the properties of control systems as they interact.

All the best,

Erling

Richard S. Marken

“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.” – William T. Powers

from Kent McClelland (2016.12.14.1600)

Rick Marken (2016.12.14.1030)

KM: A few days ago Rick offered his definitive interpretation of my “discovery,� as he called it, about collective control and conflict. Since he was referring to my work, it seems fair for me to offer a few observations of my own.

KM: Hmm … In theory, the speeed at which two systems in conflict would be expected go to their maximum output might depend on a lot of different things:

(1) How big is the difference in the references used by the interacting systems for controlling the variable in question? If it’s a big difference, the escalation will typically be fast. A small difference means a more gradual and lengthy escalation.

(2) What loop gains do the interacting systems use in controlling the variable in question? High-gain systems escalate conflicts rapidly, low-gain systems more slowly.

(3) How big are the “output reserves� that the interacting systems can draw on before reaching their maximum output? The greater the output the systems can produce, the longer the period before they hit their maximums.

(4) What other disturbances affect the variable the systems are attempting to control? If a disturbance is large enough to push the two conflicting systems to the same side of the control effort, the escalation in the divergence of output between the systems
slows down, as they “join forces� against a common enemy.

(5) What is the perceptual level of the contested variable in the conflict? The “speed of error correction,� as Bill described it (B:CP 2005, p. 67), gets slower as you go up the perceptual hierarchy. A high-level conflict, like a conflict over identities
or other system concepts, can be expected to take far longer to escalate than a low-level conflict, like arm-wrestling for instance.

(6) Are we talking about a conflict between two control systems or maybe just a few on each side of the conflict, or is it a conflict between two “giant virtual controllers,� as Martin Taylor calls them, collective control efforts involving thousands or
millions of control systems? If thousands or millions of systems are involved, and systems can enter and leave the interaction at different times, conflicts can in theory carry on indefinitely without either side reaching its maximum output. As individual
participants max out or burn out (or die), others on the same side can step in to take their place and keep the conflict going.

KM: I find Rick’s “I can’t think of anyâ€? argument—one he frequently makes when tallking about my work—kind of amusing. His inaability to think of any examples doesn’t sound like a persuasive argument to me. It sounds more like an admission of the limits of his
own thought processes and observations.

KM: As a sociologist, I’ve studied the empirical evidence about violent group conflicts for years. I’ve taught multiple courses on the topic. I could name dozens of international conflicts, civil wars, insurrections, guerrilla movements, wars against terrorism,
political struggles, racial struggles, ethnic conflicts, communal conflicts between religious groups, etc., that have lasted for years, decades, even centuries, before either side came to the limits of its output reserves. Just for one example, the FARC insurrection
in Colombia is finally being settled this month after some fifty years of conflict.

KM: Nope. Rick is the one whose arguments are being driven by his theory (a narrowly constricted view of how PCT should be done), rather than any empirical observations. His flat statement that " in real conflicts, both systems
quickly go to maximum output� is simply absurd.

KM: I’ve argued that the relative stability in all kinds of social structural and organizational patterns, such as but not limited to languages, political arrangements, economic arrangements, family patterns, religious and cultural patterns, styles in the standardized
manufacture of consumer goods, etc., etc., can be understood in the context of PCT generally, and collective control specifically. Conflict may or may not be involved in the maintenance of these slow-changing social patterns, but when it is, the stability
is always somewhat precarious, as we’ve learned in regard to political stability in the United States this fall.

KM: Rick, apparently, has had difficulty getting his head around my arguments, assuming he has actually read my papers, so if you’re interested in my work, it might make more sense to go back to the things I’ve written, instead of settling for Rick’s Cliff-Notes
version. I would be happy to share copies of my PCT-related papers with anyone who’s interested. Just contact me directly, rather than via CSGnet. The papers are also available on ResearchGate.

Best to all,

Kent

···

RM: What Kent showed is that under certain circumstances, a variable that is in conflict can appear to be controlled. Those “certain circumstances” exist when the systems in conflict are not producing maximum output. But in real conflicts, both
systems quickly go to maximum output. …

RM: So what Kent discovered is that there is a situation where a variable in conflict can be controlled, in the sense that it can be maintained in a reference state, protected from disturbance. This happens only when the systems in conflict are
not operating a maximum output, a situation that is surely very rare (I can’t think of any but perhaps you can).

RM: I still think what’s really needed are observations – preferably quantitative – of the social phenomena that you want to explain. I think the main problem (for me, anyway) with all this discussion of “collective control” is that the theory
has driven what is observed rather than vice versa.

RM: What are the social phenomena that you are trying to account for with control theory?

Erling Jorgensen (2016.12.13 1155 EST)

RM: …what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good.

RM: So saying that the variable in conflict is stabilized is the same as saying that it is controlled.

RM: No, control means stabilized in the face of disturbance that should produce instability. That’s why the stability factor is a measure of observed to expected stability. Expected stability is a measure of the variability of a variable that
would be expected on purely causal principles; observed stability is a measure of the actual variability of the variable. If the observed variability of the variable is much smaller that the expected variability it is evidence that the variable is being protected
from the effects of disturbances – ie. that it is being controlled.

RM: What Kent showed is that under certain circumstances, a variable that is in conflict can appear to be controlled. Those “certain circumstances” exist when the systems in conflict are not producing maximum output. But in real conflicts, both
systems quickly go to maximum output. When this happens, the variable no longer acts as though it’s controlled; the variable is no longer protected from disturbances. This is why powers said that when system are in conflict both lose control. It’s because
disturbances to the variable both are controlling (relative to different references) are completely effective. You can see (and experience) this loss of control in my “Cost of Conflict” demo (http://www.mindreadings.com/ControlDemo/Conflict.html).

RM: So the typical result of conflict is neither stability nor control. There will be stability, in the sense of low variability of the variable in conflict, if there are no disturbances acting on the variable. But if there are disturbances acting
on the controlled variable then there will be no stability (the variable in conflict will vary in proportion to the variable effects of the disturbance) and there will clearly be no control (the ratio of observed to expected variance of the variable in conflict
will be 1.0, indicating that the variable is not protected from disturbance at all.

RM: So what Kent discovered is that there is a situation where a variable in conflict can be controlled, in the sense that it can be maintained in a reference state, protected from disturbance. This happens only when the systems in conflict are
not operating a maximum output, a situation that is surely very rare (I can’t think of any but perhaps you can). A variable can also be stabilized by conflict, but again only in situations where there are no disturbances acting on the variable in conflict.
This may be a more common situation and it may be what Kent was talking about when he talks about conflict resulting in stability. Again, I can’t think of any socially significant examples of such situations; the only example I can think of is a tug of war
with teams of nearly equal strength. But as the tug of war situation suggests, this stability is always on the edge of instability.

[EJ] With apologies, I’ll raise a recent political example. The emerging decisions of the incoming Trump administration are very far from my control §references, & that of 65.7 millions of others, even
though they are stabilized much closer to that of 62.9 million other voters. The argument being raised in this CSGnet discussion is whether “stabilized” is a more accurate way to talk about those “virtual reference level” compromised outcomes between conflicting
control systems. In those situations, the meaning of the Stability Factor is being confounded – it is not a good measure of individual control, because of the confound of other (conflicting) control systems.

RM: I don’t see what variable is being stabilized in your political example. So I’ll just say that there is no “confounding” of the stability factor that I can see. Whether a variable is being controlled by 1 or 100 people is irrelevant to the
computation of the stability factor. All you are doing is measuring the observed variability of a variable and comparing it to the expected variability. The calculation of the expected variability does require that you know how many people are acting on the
variable and what the expected effect of those actions combined with the effect of disturbances to the variable are expected to have if the variable were not under control. But I think this can be very easily done, to quote a phase (there, two Dylan lines
in one sentence!!)

RM: I still think what’s really needed are observations – preferably quantitative – of the social phenomena that you want to explain. I think the main problem (for me, anyway) with all this discussion of “collective control” is that the theory
has driven what is observed rather than vice versa. And the social phenomenon that has been pushed as the one explained by collective control is stability. Stability is a measure of the variability of of a variable; the lower the variability of the variable,
the more stable it is. So is collective control just trying to account for the stability of social variables? Or is it trying to account for the fact that these variables are more stable than would be expected on causal grounds – that is, it is trying to
account for the fact hat these variables are controlled?

RM: What are the social phenomena that you are trying to account for with control theory?

Best

Rick

[EJ] Kent’s key argument in his recent post [Kent McClelland (2016.12.09.1400)], with which I agree, is that “we need to develop a clear analytic vocabulary for talking about the collectively controlled stabilities in our common social environments.”
I understand Bill Powers’ reluctance, and I share it, for not going beyond the bare minimum of conceptual categories in PCT. That is why I am always looking for how much mileage we can get out of the classic PCT concepts – e.g., associating conflict as mutual
“Disturbance”, noticing stable niches emerging in the “Environmental Feedback Function”, correlating various emotions with the rate of change in the “Error” term, etc.

[EJ] However, there are “cultural stabilities” that develop, like well-worn pathways, which are themselves phenomena worthy of investigation (“Phenomena Phirst!”, as the rallying cry goes). Like Kent, I’m not sure what language is best for construing
them. Various candidates have been proposed:

“Atenfel” (aesthetically, not getting my vote).

“CEV / Complex Environmental Variable” (perhaps subject to the representational misunderstanding).

“Stabilized niche” / “Well-worn path” / “Stable platform for controlling other perceptions”.

“Socially produced invariance”.

“Boss realities” that discipline reorganization.

The “Winter leaf effect” of control regions semi-protected from further disturbance.

[EJ] To speak in these terms does not keep them from also being Categories / Principles / System Concepts controlled by individuals in their personal hierarchies of control. It just allows asking broader questions about the properties of control
systems as they interact.

All the best,

Erling

Richard S. Marken

“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.” – William T. Powers

[Martin Taylor 2016.12.17.23.07]

[From Erling Jorgensen (2016.12.14 0945 EST)]

>[Martin Taylor 2016.12.13.14.32]
>>Erling Jorgensen (2016.12.13 1155 EST)]

>[MT] I don't like the term "Stability Factor" because a lot of uncontrolled
things are stable. Rick mentioned El Capitan. I might add a weight hung
by a spring or a ball in a bowl. The term I prefer is "Control Ratio"
for the ratio Rick describes. That's teh wording that used to be
standard on CSGnet.
[EJ] I don't recall the use of that term, "Control Ratio," but I like it. Perhaps it was used in the earlier days of CSGnet, before I had joined the listserv.

In <http://www.mmtaylor.net/PCT/Info.theory.in.control/Control+correl.html>, which was a 1998 posting to CSGnet, I used "Control Ratio" to refer to an amplitude ratio rather than a variance ratio between controlled and uncontrolled variation under the influence of a disturbance. I guess that I must have done so because that was the conventional usage at the time.

  I can also understand a judgment call about how "Stability Factor" might be intriguing to a non-PCT readership, without eliciting [sorry about the behavioral framing here] the knee-jerk reaction against the notion of "control."

>> [EJ] In a tug-of-war with each team pulling as hard as they can, neither
>> team has yet "controlled" the flag by bringing it to their preferred
>> position, if the flag is still hovering ("stabilized") in the middle.
>> Bill Powers' himself said -- (I love Rupert's affectionate
>> designation, the Big Powowski!) -- that control systems in intractable
>> conflict have effectively _lost_ control.

>[MT] The individual controllers may have lost control, but that does not mean
that the position of the "flag" must be uncontrolled. In a tug-of-war,
it actually is uncontrolled because both teams are pulling as hard as
they can, but in other analogous situation in which they aren't, the
"flag" appears to be controlled by a system with a gain that is the sum
of the two gains, and a reference value given by the ratio of the gains
between the two individual reference values. [...the quotation continues further below...]
[EJ] I believe that is the issue being discussed here: I.e., whether "control" is the best word for those interactive situations where there is the appearance of control, but the relevant perception is not sufficiently close to either party's actual reference. I'm arguing that "stabilized" is better, to emphasize that the outcome may derive from competing systems whose actual "control" is pretty poor, where control refers to tracking their own respective reference level. This doesn't mean that we can't still use the whole PCT armamentarium to analyze the situation. (Notice, as well, that "stabilized" as an verb acting in the situation seems better than "stability" as a property of the system.)
[EJ] Perhaps we need to resurrect the old language of "controlling for" a given value of a perception.

I must be getting old. I have continued to use "controlling for" as a short form of "Controlling a perception of with a reference value of". For example. I "control for the room temperature to be 19C" is a short form of "I control a perception of the room temperature, with a reference value of 19C".

  Maybe we should consider a term such as "agent based virtual control," similar to the work with Agent-Based Computational Economics that Charlotte Bruun has done.
[EJ] Returning to Martin's interrupted quotation above...
>[MT] Unless the tester can
actually see the actors, there's no way that the TCV could tell the
difference between what I call a "Giant Virtual Controller" and a real
physical controller.
[EJ] I wonder about this. I'm thinking spatially, & considering the simplified case of just two competing control systems pulling in opposite directions. And I guess it may be important that their respective outputs have not yet maxed out. I'm picturing it as two adjoining circles, representing the two control systems, where the radius represents each one's output pulling on a common variable, which is stabilized on a tangent point between the two circles.
[EJ] There are various directions an additional Disturbance could be applied in the Test for the Controlled Variable. If the TCV disturbance is strictly orthogonal to both systems (i.e., pulling sideways out the joint tangent line), the output of both systems would be predicted to increase, because the variable gets further away from each of their preferred reference for it. But what if the TCV disturbance is applied along the radius of one of the circles, or in some other manner. Wouldn't there be a difference in their respective outputs, & could that differential help to distinguish between a single controller being involved versus a virtual controller with more than one control system contributing to the outcome?

I'm not going to try to answer this, except to say that I agree with you and am (and for some days have been) writing a message describing a gedanken experiment that I hope will help people to think about most of the issue you raise or continue in this message.

One concept I would like people to think about before I send that message is "neural current". No neural current exists in the brain. It's a virtual property of events happening in a whole chunk of the brain (a neural fibre bundle according to WTP).

[EJ] For me, it's not just aesthetics, but rhetorically, too. While I haven't read the LCS-IV draft contributions where I believe the full argument is laid out, I've tried to notice my own reactions when you or Kent have used that term "atenfel." I always notice myself slowing down my absorption of the train of thought, while I tried to figure out 'What were those words again that the letters represent?' In that sense, "atenfel" gets in its own way. It doesn't easily build on an evocative or associative context, to get me part of the way there, the way the word "cybernetics" could possibly evoke "steersmanship." So, for now, I notice myself inserting a term such as "feedback path" when I run into the word "atenfel."

That's unfortunate, because it is likely to mislead you to do so. Think more of "tool", which has connotations more similar to "atenfel". A "tool" is used in the control of a perception, or at least some property of the tool is so used. "Environmental Feedback Path Segment" is too passive, as if it was just a piece of wire, when it might be the workings of a complex system, say a bank that provides services that help you to get food on the table. (Not that you couldn't do it another way that doesn't involve a bank). Or it might be simple, such as the sharpness of a pair of cutters that you have the skill to use in a supporting perceptual control loop when you control for pruning a tree. The entire control loop that involves the observable action of wielding the cutters is an atenfel in your controlling for perceiving the tree-branch to be severed.

>[MT] It may be the result of an environmental
stability or a stability created by individual or collective control,
but an atenfel is just a segment of the environmental feedback path
between the output of an ECU (Elementary Control Unit -- another term)
and the perceptual input of that same ECU.
[EJ] So, how about the term "EFF-segment"? Environmental Feedback Function is already in the vernacular, even though it is not emphasized as much as the other functions in the control loop. I think 'segment' is the additional point you've now clarified for me, in how you use the concept.

Segment, yes, but a segment of a particular kind, functional.

>[MT] It's in a conceptually
different domain, just as a wire is in a conceptually different domain
from a voltage.
[EJ] Yes, this is the distinction I was trying to make by distinguishing the "properties" of the loop from the "values" that travel along the wires.

>> [EJ] "CEV / Complex Environmental Variable" (perhaps subject to the
>> representational misunderstanding).

>[MT] A CEV is not a cultural stability. It's just a variable that is a
function of "simpler" variables. When a similar concept refers to a
cultural stability (i.e. a collectively controlled value of an
environmental function) I call it a CCEV (Collective Complex
Environmental Variable, to distinguish the two concepts.
[EJ] I appreciate hearing how you use & define these concepts. I don't believe there is yet consensual agreement among the PCT/CSG community on the use of these terms.

No,there isn't. Some even seem to be controlling for the very concepts to be expunged from discussions of PCT. And CCEV is a concept that I hope my in-progress message will elucidate. Since I invented both terms, I think I may carry some weight in describing their intended usages.

>[MT] A CCEV corresponds to a virtual perception controlled by a Giant Virtual
Controller to some reference value that may not be held by any of the
individual controllers whose controlling results in the observed
effects.
[EJ] ... Perhaps the key language we are seeking is in that word "virtual." Speaking of a "virtual reference level" for a "virtual perception" in a "Giant Virtual Controller" at least makes clear that the resulting stability may or may not line up with any individual's preferred reference level in what they are attempting to control.

My reference above to "neural current" is intended to suggest that this may (perhaps must) also be true within an individual brain.

[EJ] I believe there is pretty good convergence between us on what we are trying to talk about here. Just some disagreement still on the best terms to use. ..
[EJ] I think that's what we're doing here -- trying to work out a clear & useful "analytic vocabulary for talking about the collectively controlled stabilities," to use Kent's phrase.
[EJ] To the above list of associated concepts, I think we've now added the following, (admittedly, this is a clustering of ideas that may not all have the same referent):

To my mind they don't.

"Controlling for"

A shorthand way of saying "controlling perception P with a reference value V"

"Agent based virtual control"

No idea what that might mean

"Giant Virtual Controller"

The impression that something is controlling an observable environmental property that occurs when a number of individual controllers control related perceptions, whether their reference values are similar or wildly different.

"Feedback path"

Everything that occurs between the output of an elementary control unit (ECU -- another term in the dictionary) and its perceptual input function.

"EFF-segment"

Any part of an environmental feedback path (which is another reason why it does not correspond to "atenfel", since an atenfel is a functional component of the environmental feedback path, not just any segment).

[EJ] I see this whole discussion as an _example_ of collectively controlled convergence as to perceptions, (well, hopefully convergence), in addition to an _analytically_ useful conversation.

That would be nice, indeed. I (and I think Kent) believe this is a way that languages evolve. Another way is simply to use the words consistently in contexts where their meanings are reasonably clear, so that those meanings can be used in context from which the meanings cannot be inferred so clearly. I remember in my early youth having used "fatal" as an adjective to describe a minor accident to my mother, and being informed that "fatal" couldn't be used in that context. I tried something similar in respect of "atenfel" above. I hope we can converge on an understanding of the concept, and perhaps on a more mellifluous word for it. Remember that the word is just one of a related suite "atenex", "molenfel", and "molenex". I hope that a better replacement word for "atenfel" also carries along its relations.

Martin

···

All the best,
Erling

[From Rick Marken (2016.12.17.2200)]

···

Kent McClelland (2016.12.14.1600)–

KM: A few days ago Rick offered his definitive interpretation of my “discovery,â€? as he called it, about collective control and conflict. Since he was referring to my work, it seems fair for me to offer a few observations of my own. Â

KM: Hmm … In theory, the speed at which two systems in conflict wouuld be expected go to their maximum output might depend on a lot of different things:Â

RM: You are so right. Indeed, systems in conflict don’t necessarily go to maximum output ever. I based my statement about the systems going “quickly” to maximum output on a little conflict simulation that I had set up. I just checked it out in preparation for answering this post and found that I had made a rather basic error; I had the output of only one system (rather than both) affecting the controlled variable. When set up properly, what I found is exactly what you found in your simulations; that a variable being controlled relative to different reference levels by different control systems will be held in a reference state (relative to a virtual reference specification with a value between the values of the reference specifications of the two systems in conflict), protected from disturbances. That is, it will be controlled as long as disturbances don’t force the systems involved to maximum output, at which point control is lost.

KM: I find Rick’s “I can’t think of anyâ€? argument—one he frequently makes when talking about my work—k€”kind of amusing. His inability to think of any examples doesn’t sound like a persuasive argument to me. It sounds more like an admission of the limits of his
own thought processes and observations.Â

RM: Yes, my mistake. I thought the systems involved in the conflicts I am familiar with – tug of war, my on line demo – were operating at maximum output; but this is not necessarily the case. What I should have said is that I can’t think of any examples of social variables that are being controlled (they are not just stabilized since they are protected from the effects of disturbance) because two or more control systems are trying to maintain that variable in different reference states. I’m not saying that such controlled social variables – controlled as a result of conflict –  don’t exist. Â

KM: As a sociologist, I’ve studied the empirical evidence about violent group conflicts for years. I’ve taught multiple courses on the topic. I could name dozens of international conflicts, civil wars, insurrections, guerrilla movements, wars against terrorism,
political struggles, racial struggles, ethnic conflicts, communal conflicts between religious groups, etc., that have lasted for years, decades, even centuries, before either side came to the limits of its output reserves. Just for one example, the FARC insurrection
in Colombia is finally being settled this month after some fifty years of conflict. Â Â

RM: Yes, but what I would like to know is what is the variable that is being controlled (kept at a virtual reference level) in each of these conflicts.

KM: I’ve argued that the relative stability in all kinds of social structural and organizational patterns, such as but not limited to languages, political arrangements, economic arrangements, family patterns, religious and cultural patterns, styles in the standardized
manufacture of consumer goods, etc., etc., can be understood in the context of PCT generally, and collective control specifically.

RM: Yes, but what I would like to know is exactly what variable aspects of language, political arrangements, family patterns, etc, are stabilized by conflict. And how is the control effected. What are the outputs of the actors engaged in the conflict that affect this aspect of language? What are the disturbances to this aspect of the language that are being opposed by these outputs?Â

Â

KM: Rick, apparently, has had difficulty getting his head around my arguments,

RM: I have read your papers and I only have difficulty going from your nice simulations of conflict to mapping how these simulations map to the “stability” of the things you talk about: languages, political arrangements, economic arrangements, family patterns, religious and cultural patterns, styles in the standardized manufacture of consumer goods, etc. It’s not that I don’t believe that your model can account for the stability of variable aspects of these phenomena, it’s just that I would like to see an explicit application of the model, showing how the variables in the model correspond to the variables observed in these social phenomena.

BestÂ

Rick

Â

assuming he has actually read my papers, so if you’re interested in my work, it might make more sense to go back to the things I’ve written, instead of settling for Rick’s Cliff-Notes
version. I would be happy to share copies of my PCT-related papers with anyone who’s interested. Just contact me directly, rather than via CSGnet. The papers are also available on ResearchGate.Â

Best to all,

Kent

On Dec 14, 2016, at 12:33 PM, Richard Marken rsmarken@gmail.com wrote:

[From Rick Marken (2016.12.14.1030)]


Richard S. MarkenÂ

“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.” – William T. Powers

RM: What Kent showed is that under certain circumstances, a variable that is in conflict can appear to be controlled. Those “certain circumstances” exist when the systems in conflict are not producing maximum output. But in real conflicts, both
systems quickly go to maximum output. …

RM: So what Kent discovered is that there is a situation where a variable in conflict can be controlled, in the sense that it can be maintained in a reference state, protected from disturbance. This happens only when the systems in conflict are
not operating a maximum output, a situation that is surely very rare (I can’t think of any but perhaps you can).

RM: I still think what’s really needed are observations – preferably quantitative – of the social phenomena that you want to explain. I think the main problem (for me, anyway) with all this discussion of “collective control” is that the theory
has driven what is observed rather than vice versa. Â

RM: What are the social phenomena that you are trying to account for with control theory?

Erling Jorgensen (2016.12.13 1155 EST)

RM: …what Kent showed in his modeling work is that, in conflict situations, the variable in conflict is controlled: the ratio of actual to expected variance is very small so stability is high meaning control is good. Â

Â

RM:Â So saying that the variable in conflict is stabilized is the same as saying that it is controlled.

 RM: No, control means stabilized in the face of disturbance that should produce instability. That’s why the stability factor is a measure of observed to expected stability. Expected stability is a measure of the variability of a variable that
would be expected on purely causal principles; observed stability is a measure of the actual variability of the variable. If the observed variability of the variable is much smaller that the expected variability it is evidence that the variable is being protected
from the effects of disturbances – ie. that it is being controlled.Â

RM: What Kent showed is that under certain circumstances, a variable that is in conflict can appear to be controlled. Those “certain circumstances” exist when the systems in conflict are not producing maximum output. But in real conflicts, both
systems quickly go to maximum output. When this happens, the variable no longer acts as though it’s controlled; the variable is no longer protected from disturbances. This is why powers said that when system are in conflict both lose control. It’s because
disturbances to the variable both are controlling (relative to different references) are completely effective. You can see (and experience) this loss of control in my “Cost of Conflict” demo (http://www.mindreadings.com/ControlDemo/Conflict.html).Â

RM: So the typical result of conflict is neither stability nor control. There will be stability, in the sense of low variability of the variable in conflict, if there are no disturbances acting on the variable. But if there are disturbances acting
on the controlled variable then there will be no stability (the variable in conflict will vary in proportion to the variable effects of the disturbance) and there will clearly be no control (the ratio of observed to expected variance of the variable in conflict
will be 1.0, indicating that the variable is not protected from disturbance at all.Â

RM: So what Kent discovered is that there is a situation where a variable in conflict can be controlled, in the sense that it can be maintained in a reference state, protected from disturbance. This happens only when the systems in conflict are
not operating a maximum output, a situation that is surely very rare (I can’t think of any but perhaps you can). A variable can also be stabilized by conflict, but again only in situations where there are no disturbances acting on the variable in conflict.
This may be a more common situation and it may be what Kent was talking about when he talks about conflict resulting in stability. Again, I can’t think of any socially significant examples of such situations; the only example I can think of is a tug of war
with teams of nearly equal strength. But as the tug of war situation suggests, this stability is always on the edge of instability.Â

[EJ] With apologies, I’ll raise a recent political example. The emerging decisions of the incoming Trump administration are very far from my control §references, & that of 65.7 millions of others, even
though they are stabilized much closer to that of 62.9 million other voters. The argument being raised in this CSGnet discussion is whether “stabilized” is a more accurate way to talk about those “virtual reference level” compromised outcomes between conflicting
control systems. In those situations, the meaning of the Stability Factor is being confounded – it is not a good measure of individual control, because of the confound of other (conflicting) control systems.Â

RM: I don’t see what variable is being stabilized in your political example. So I’ll just say that there is no “confounding” of the stability factor that I can see. Whether a variable is being controlled by 1 or 100 people is irrelevant to the
computation of the stability factor. All you are doing is measuring the observed variability of a variable and comparing it to the expected variability. The calculation of the expected variability does require that you know how many people are acting on the
variable and what the expected effect of those actions combined with the effect of disturbances to the variable are expected to have if the variable were not under control. But I think this can be very easily done, to quote a phase (there, two Dylan lines
in one sentence!!)Â

RM: I still think what’s really needed are observations – preferably quantitative – of the social phenomena that you want to explain. I think the main problem (for me, anyway) with all this discussion of “collective control” is that the theory
has driven what is observed rather than vice versa. And the social phenomenon that has been pushed as the one explained by collective control is stability. Stability is a measure of the variability of of a variable; the lower the variability of the variable,
the more stable it is. So is collective control just trying to account for the stability of social variables? Or is it trying to account for the fact that these variables are more stable than would  be expected on causal grounds – that is, it is trying to
account for the fact hat these variables are controlled?Â

RM: What are the social phenomena that you are trying to account for with control theory?

Best

Rick

Â

[EJ]Â Kent’s key argument in his recent post [Kent McClelland (2016.12.09.1400)], with which I agree, is that "we need to develop a clear analytic vocabulary for talking about the collectively controlled stabilities in our common social environments."Â
I understand Bill Powers’ reluctance, and I share it, for not going beyond the bare minimum of conceptual categories in PCT. That is why I am always looking for how much mileage we can get out of the classic PCT concepts – e.g., associating conflict as mutual
“Disturbance”, noticing stable niches emerging in the “Environmental Feedback Function”, correlating various emotions with the rate of change in the “Error” term, etc. Â

Â

[EJ] However, there are “cultural stabilities” that develop, like well-worn pathways, which are themselves phenomena worthy of investigation (“Phenomena Phirst!”, as the rallying cry goes). Like Kent, I’m not sure what language is best for construing
them.  Various candidates have been proposed: Â

“Atenfel” (aesthetically, not getting my vote). Â

“CEV / Complex Environmental Variable” (perhaps subject to the representational misunderstanding). Â

“Stabilized niche” / “Well-worn path” / “Stable platform for controlling other perceptions”. Â

“Socially produced invariance”. Â

“Boss realities” that discipline reorganization. Â

The “Winter leaf effect” of control regions semi-protected from further disturbance. Â

Â

[EJ] To speak in these terms does not keep them from also being Categories / Principles / System Concepts controlled by individuals in their personal hierarchies of control. It just allows asking broader questions about the properties of control
systems as they interact. Â

Â

All the best,Â

ErlingÂ

Â

–Â
Richard S. MarkenÂ

“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.” – William T. Powers