Plant volatiles are multi-species CCVs

Here is a very interesting investigation of variables collectively controlled by diverse species.

Yoneya, K. and Miki, T. (2015), Co-evolution of foraging behaviour in herbivores and their natural enemies predicts multifunctionality of herbivore-induced plant volatiles. Funct Ecol, 29: 451-461.

Collectively controlled variables evolve cross-generationally at a rate of change that is generally too slow for individuals to perceive, and at a given time are established properties of the environmental feedback paths available to progeny as well as adults.

The only human participation considered is implicitly that of the investigators. As is typical, it is as though ‘the ecosystem’ were a separate ‘terrarium’ system monitored from the outside.

A bit tangential: advocates of Paleo and Carnivore diets sometimes characterize the defensive mechanisms of plants as a ‘warfare’ between plants and animals which humans should abandon. Even Paul The Carnivore Code Saladino has changed his mind. New Yorker: Is an all-meat diet what nature intended?.

There are many processes, where “…diverse effects resolve to a single result”, that don’t involve control. One of them is the process by which the cell body of a neuron resolves the diverse pulse trains entering it via dendrites into the single pulse train leaving it via the axon. I think we need a better definition of collective control. Doesn’t collective control have something to do with virtual controlled variables?

I just read the abstract and it sounds very interesting. But it would help me understand why you think collective control is involved in this co-evolutionary process if you would describe what you think are the collectively controlled variables and explain why you think they are controlled, let alone collectively controlled.

Such is the skeptical attitude of PCT.

If it can be shown that this results from the control actions of distinct control systems within the cell body of a neuron then the definition is good enough. Likewise, if no one has discovered autonomous control systems interacting within the neural cell with this collective result (and that is the case) the example has no claim on the definition, and the definition is good enough.

Yes, that is the “single result”.

If you are in fact interested, you will read more than the abstract and look for examples and maybe data; if you are not, you won’t.

You have organisms of three kinds: forageable plants, foraging herbivores that eat from them, and predators that eat the herbivores.

Herbivore-induced plant volatiles (HIPVs) are volatile substances that plants produce when they are eaten and which diffuse widely in the air.

The blend of HIPVs is varied by the plant depending on

  • current physiological status, including the degree of damage and its maturational stage
  • ecological status, including what herbivores are eating them.

“In addition, high volatile chemical diffusivity enables multiple species inhabiting plants and the surrounding environment to share information …. These specific infochemical features allow HIPVs to act as multifunctional signals in a community, for example attracting … and repelling … herbivores, and attracting … and repelling … their natural enemies.”

Herbivores use these to control a number of variables:

  • abundance of consumable plants
  • nutritional quality of plants
  • risk of predation.

For herbivores, it easier to find plants that have already suffered grazing (more HIPVs) but plant health and therefore nutritional quality suffers.

The same HIPVs are CVs of predators.

For predators, HIPVs from foraged plants are inputs to PIFs for prey, but prey in over-grazed places are less healthy and therefore nutritional quality suffers.

Their computational model accounts for the data “understood as combination of evolutionarily stable strategies (ESSs) of the herbivore and its natural enemy.” These strategies are examples of collective control. Given the animals involved, I would guess the peripheral and internal loop functions are genetically determined, but with the tremendous growth in understanding of plasticity and adaptation there is no longer an impervious conceptual wall between learning and evolution.

Reasonable, I think, to expect such strategies shaping vertebrate and mammalian predator-prey relations around the same HIPVs, and that pheromones of predators and prey may well also be involved.

As usual, I don’t quite understand what you are saying. But I think you might be saying that we can’t say that seeing "diverse effects resolve to a single result” is “collective control” unless we can show that the entities producing the diverse effects are control systems. If you do this analysis I think you’ll find that some of the most interesting cases of what might appear to be collective control do not involve any control systems at all. An example of such a phenomenon is spontaneous synchronization (SS).

There are actually many different examples of SS, all of which are lovely examples of diverse effects resolving to a single result. Several examples of SS are described in this video. Some of these could be called collective control but others are simply results that are produced by a collective of causal systems; perhaps they should be called examples of collective causality. I suggest you watch the video and see if you can tell which phenomena are examples of collective control and which are examples of collective causality.

I, therefore, propose that your definition of collective control as "a phenomenon where diverse effects resolve to a single result” be augmented to read "a phenomenon where diverse effects produced by a set of control systems resolve to a single result”, and that we use your definition of collective control as the definition collective causality.

The phrase “the participating autonomous control systems” does say that control systems are producing the effects, on any reasonable reading of “participate”. (With more context: the phrase “…whether the participating autonomous control systems are cells or …” etc.) It is good to make that more explicit, but I think you’ve overshot the mark, because not all the effects on the CCV are intended. We have to admit there can be unintended side effects of control and environmental disturbances from other than controllers.

Although not as explicit as we might like, it does demarcate a population of autonomous control systems from a “collective of causal systems” which are not control systems. Note that collective control can be maintained in an environment that includes both, the latter as sources of environmental disturbances.

In the CROWD demo, add a third kind of agent. Metaphorically imagine a seller of ice cream with loop functions (PIF, comparator, reference, OF) for putting its wares near crowds. The existing agents, or some of them would have to have loop functions for acquiring those wares. For that agent, the arcs and rings are collectively controlled environmental phenomena which it is using as part of its environmental feedback function. Over to you.

How about we preface it with “Collective control can occur in a population of autonomous control systems.”

Here’s what I think. I think the term “collective control” should be eliminated from the PCT vocabulary. It’s just confusing. I’ve been reading a draft of a 2010 paper by Kent entitled “Collective Control and Environmental Stabilization”. In that paper Kent defines “collective control” as follows:

However, when the actors involved in the joint control of an environmental variable use different reference points in attempting to control the variable, their actions will come into conflict, because the outcome of this “collective control process, as it has been called (McClelland 2006), will stabilize the controlled variable at a “virtual reference level” (Powers 2005: 267).

So, “collective control” is a result of CONFLICT. Kent goes on to say “From the perspective of PCT, every kind of social process is a process of collective control” [emphasis mine]. But he then gives examples of collective control where no conflict is involved at all:

…a choir hum[ing] a note together at a single frequency" and "a movie video, as when a film production company combines visual images, sounds, actors’ performances, voice-over narration, and computer effects to construct a reproducible artifact…

The stabilized results produced in each of these cases – simultaneously sung notes of the same pitch or a video – are not virtual reference states resulting from conflict; they are real reference states resulting from control

Since the meaning of “collective control” is frustratingly unclear, I suggest dispensing with it and just using the term “social control” to refer to the behavior produced by groups of control systems. This is the term used to describe such behavior in my book The Study of Living Control Systems.

In that book I distinguish three different types of social control observed in collections of control systems: 1) cooperative control, which can be intended (as in industrial production) or unintended (such as the flocking patterns produced by birds), 2) conflictive control (which isn’t really control at all, as per my talk at the 2023 IAPCT conference), and 3) manipulative control.

In my treatment of social control I show how these different types of social behavior emerge from the controlling done by the individuals in a group. These different instances of social control don’t all emerge from the same process; but they all emerge from the controlling done by the individuals in the collective. There is no “giant virtual controller” making these social phenomena happen. At least I don’t see any reason to think so and neither did Bill or Tom Bourbon.

Yes, I have seen that you are unable to perceive social phenomena in this way.

Kent gave an example of collective control resulting from conflict. From this, you have concluded “So, “collective control” is a result of CONFLICT”. You find it “frustratingly unclear” that Kent also gives examples of collective control that do not involve conflict.

Kent’s simplest examples which he first presented over 30 years ago show how ‘stabilization of an environmental variable’ can result from conflict. The problem, my friend, is in your demand that collective control can only take this form.

There is nothing unclear about it. Stabilization due to conflict is a limited corner case within the larger body of collective control phenomena.

Your suggestion that we use “social” instead of “collective” to name the phenomena is no more than a change of terminology with zero substance. I’m sorry, but you don’t get to change the terminology solo. Others already have some investment in this class of phenomena being identified as ‘collective control’. In other words, uncomfortable though saying it may be, the association of the name with the phenomena is a perceptual variable which is collectively controlled by participants in the field, of whom you are only one. You may push on that variable, but the collective effect of others resisting your push is very likely to be stronger.

You are experiencing effects of collective control when, having demanded that collective control must be understood as always and only resulting from conflict, you get replies from others who have investigated the subject telling you that your restricted definition of collective control is mistaken and affirming a broader definition on which they agree.

No one asserts that a “giant virtual controller” exists in the way that the individual people exist. It exists only as a theoretical entity in exactly the same way that ‘neural bundle’ and ‘neural current’ exist. In B:CP (p. 22) a ‘neural current’ is “the number of impulses passing through a cross section of all parallel redundant fibers in a given [neural] bundle per unit time”. These are theoretical entities abstracted from complex and messy facts of neuroanatomy and neural function.

We need these simplified theoretical entities to encapsulate complexity because otherwise the complexity of detail overwhelms the possibility of useful research and results.

A neuroscientist or neuroanatomist looking at our control loop diagrams is well justified in doubting their simplicity. We can admit yes it is all much more complicated, but we predict that in some way the complexity will amount to functions of these kinds connected in these ways, and we are confident of that because the observed result is Control, and there is no other known structure that results in control.

The neurological and collective simplifications are exactly parallel, because in the collective situation the observed result is Control, even though no one individual participant is doing all of that controlling.

A functional block diagram of the brain's organization, however, can remain valid even if it were later to be discovered that each function is a distributed property of the brain and each perception is a pattern of neural currents pervading the brain. (*B:CP* 38)

A functional block diagram of collective control can remain valid even when we know that each function of the posited Giant Virtual Controller is a distributed property of individual autonomous control systems in a population.

Not everyone has the interest or perhaps even the perceptual input functions to take an interest in the complexities of social phenomena. Those of us who do, need the theoretical and methodological tools to grapple with their complexity in an orderly way. Each theoretical entity must be reducible in principle to the simpler terms of the more basic aspects of our science, just as ‘neural bundle’ and ‘neural current’ must have a plausible relationship to the anatomical and neurodynamic data with which a neuroscientist must grapple. You have no more basis for dismissing collective control than a neuroanatomist has for dismissing PCT.

Yes, I am unable to see many (actually, most) social phenomena in terms of what you call “collective control”.

Yes, you are right. While the collective control model is mainly about conflict-based phenomena, I now see that it can also be a model of cooperative phenomena. After re-reading Kent’s “Collective Control and Environmental Stabilization” paper more carefully I see that “collective control” refers to a model of social behavior where a collection of controllers are all controlling the same variable. If these controllers are all controlling that variable relative to the same reference the result is cooperation. But, as Kent notes in his paper, this situation is “highly unlikely”. So the “collective control” model is mainly an explanation of social “stability” resulting from conflict, where the conflict maintains a commonly “controlled” variable (CCV) in a virtual reference state.

No, I found the term “collective control” frustratingly unclear. But now I think it’s meaning is clear (to me, anyway): it’s refers to a model of social processes as a collection of controllers all controlling the same variable. According to Kent, every kind of social process is a process of collective control. So “collective control” refers to a general model of “every kind of social process”.

But I think Kent’s claim that the “collective control” model can account for every social process is demonstrably wrong. In The Study of Living Control Systems (SLCS) I present examples of several social processes – two-person cooperation, flocking, crowd behavior, social differences in pronunciation – that are accounted for by models that are nothing like the “collective control” model.

And I certainly didn’t “demand” that collective control can only take the form of a conflict-based model. I had thought the model was very limited because it was conflict-based but now I see that it can account for some cooperation, but, as Kent admits, this application of the model is rare since it is unlikely that all members of a collective of agents will adopt exactly the same reference for the variable they all want to control.

But as I just noted, Kent said that non-conflictual instances of “collective control” are bound to be very rare. This means that stabilization due to cooperation is a very limited corner case within the much larger body of collective control phenomena which are based on conflict.

My chapter on “Social Control” in SLCS shows the substantive changes in one’s approach to modeling social phenomena that can come about by abandoning the term “collective control”, which would also involve abandoning adherence to the idea that the “collective control” model is a general model of social behavior.

And it is the wrong theory! The “virtual controlling” done by the theoretical “giant virtual controller” is nothing like the controlling done by an actual control system.

Good. But there is more to read.

Consider how perceived features of the environment may serve as a segment within the environmental feedback path for controlling perceptions which are otherwise entirely unrelated to those features. Diverse environmental feedback paths include that segment.

A deer trail in the forest passes between a boulder on the north and a large tree and thorn thicket on the south. Deer pass that way to and from water. Human hikers pass that way to and from a scenic overlook. At a stage of their life cycle ticks climb out on bordering huckleberry leaves and endeavor to attach themselves to passing deer, humans, raccoons, etc.

Consider the built environment in which you live coddled by anonymous others who maintain it. Not far from your front door is a paved road which connects to other paved roads. Along the center of such a roadway is perhaps a painted line, and perhaps others painted along its borders. Sometimes the center line is double, sometimes broken into dashes, sometimes a combination of solid and dashed. Sometimes the paved surface breaks and the opening breaks down to a growing pothole. These are a few among the countless collectively controlled perceptions that constitute our built environment.

I have not considered it because it makes no sense. When is features of a segment of a feedback path – or the feedback path as a whole – ever related to controlled perceptions? Let’s take the feedback path between mouse and cursor position as an example. The feedback path goes from hand movement to mouse movement to a Bluetooth carrier to a computer memory cell to the display. There is nothing in that path that is related to the controlled perception (distance between cursor and target) other than by being part of a causal chain that relates hand movement (system output, o) to cursor-target distance (controlled perception, p). When modeling this process the complex feedback path from o to p can be written as p = k*o.

Sure, possibly. For example, when you run my Basic Tracking Task on your computer, the feedback path from your movement to cursor movement is surely somewhat different than when I run it on mine; you might use a joystick rather than a mouse, have a different operating system so that the software that converts your movement into cursor movement is somewhat different, and so on. But there certainly may be segments that are the same.

The trail is a possible feedback path between the output and controlled variable of the deer or hikers. But you have to know what variable is being controlled and what outputs are used to control it in order to see how the trail functions as the environmental feedback portion of the loop. The same trail could function quite differently for, say, deer controlling for finding the water hole versus the hikers controlling for getting to the lake before sundown.

But I don’t see how the existence of human or animal produced “feedback paths” addresses my claim that Kent was demonstrably wrong when he said that “the “collective control” model can account for every social process”. I think it can account for tug of wars and wars in general. But that’s pretty much it. It can’t account for most cooperative social processes such as string quartets and symphonies, iPhones and Titan rockets, etc.

On one of those nearby paved roads that you travel a crosswalk is painted. And occasionally repainted.

You are engaged in synecdoche, taking a part to be the whole. What you refer to as “the ‘collective control’ model” is a few careful models of the simplest forms of collective control. Baby steps.

In fairness, I think Kent (at least as you quote him) has it backward. PCT must account for every social process. Social processes inherently involve plural agents controlling in a shared environment. That environment includes variables which are controlled (when disturbed they are restored to an observable reference value) but not entirely due to the control activities of any one of the individual agents. They may trade off periods of active control (“Ah, right, I scheduled myself to distribute bags from the Food Pantry this Wednesday, rain check?”), delegate or invoke delegation of active control (“Yeah, I called, and they said the crew working over on Potter would swing by and repaint that crosswalk.”), and so on.

Goffman was a close observer of collectively controlled variables in public places, as for example how pedestrians on a busy sidewalk indicate intent to oncomers and avoid collisions (angle of foot, of hand, quick glance then direction of gaze, slant of shoulder, microsecond pause, lowering bend of knee during a stride, and so on and on). Relations in Public: Microstudies of the Public Order is one good trove. I’m sorry I never got to know him. Dell Hymes was in process if persuading him to come to Penn just as I was leaving for fieldwork in 1970.

I’m afraid I do not have the pleasure of understanding you. What is the controlled perception? What is the output that is connected to the controlled perception by the feedback path. What is the feedback path? What is the segment of the feedback path? And what is the feature of this segment of the feedback path that is related to the controlled perception?

So, now the “collective control” model is not a model of collective control? It’s just a model of the simplest forms of collective control? I suggest, for clarity, that you change the name of the model to the name of each simple form of collective control to which it applies. Something like what Bill and I did with our models of examples of social behavior (the phenomenon previously known as “collective control”). Bill developed a wonderful model of crowd behavior and he called it the CROWD model. I developed a model of pronunciation drift in different sub-populations of speakers and I called it the “pronunciation drift model”. I think if you guys did that – name your models after the phenomena they explain – it would help clarify what you actually know about social behavior.

Exactly. So why don’t you guys help me (and others) avoid the syn of synecdoche and hold off on claiming you have a model of collective control until you really have a model of collective control.

I took sociology classes from a very young Harvey Sacks when I was an undergrad at Ucla (he was already a legend in sociology circles when he died in a car crash only a few years after I took courses from him). Sacks, like Goffman, was a great observer of social processes and he understood (intuitively, not explicitly) that they were based on control processes (that they were purposeful). One of my favorite observations of his was that young kids typically say “You know what?” when they want to say something to an adult. He saw that the kids did this because they were controlling for getting the adult to ask them a question – What? – because they have learned that adults are more likely to listen to you when you are answering their questions.

So can we stop calling Kent’s model a model of “collective control” and just call each version of the model by the name of the phenomenon it explains? That would be very helpful.