Controlled Variables are Perceptual Variables

Fundamentals
5

[Martin Taylor 2020.10.22.12.20]

This is a response both to Rick and to a CSGnet posting by Warren Mansell.

Here’s the relevant segment from Rick’s relevant posting to CSGnet. Note his substitution of “external reality” for my “Real Reality”.: On Sun, Oct 25, 2020 at 7:48 AM Martin Taylor <csgnet@lists.illinois.edu> wrote:

Rick forgets that one’s actions influence real reality, whereas one perceives (and creates thereby) a perceived reality from the resulting input quantity. FN, WM, and EJ realize this and their quoted statements are correct. The precision with which the perceptual variable’s Perceived Reality corresponds to that of a variable in Real Reality determines the best possible Quality of Control. Real Reality may not function anything like Perceptual Reality, but the effects we produce on Real Reality do affect the input quantities that correspond to variables we perceive and control.

RM: No, I didn’t forget that actions influence external reality. See my reply to Bruce Nevin at the IAPCT Discourse site.

The rest is my response to Warren’s “On 2020/10/21 11:22 PM, Warren Mansell (wmansell@gmail.com via csgnet Mailing List) wrote”, but it applies equally to Rick’s unfortunate sunstitution of “external” for “real”.

All we control is our perception of the variable; that doesn’t mean that the variable itself isn’t also controlled, but we have no way of directly knowing that because we are all perceptual control devices even the people with rulers and electronic measuring devices. We just have to assume it is the case because we’ve survived and haven’t crashed our car yet…

I think it is a little more assured than that, though that is indeed part of the argument.

The argument I would use is that we perceive and directly control a variable property of something in Perceived Reality (PR), but can act only on whatever is really “out there”, a Real Reality (RR) variable. Disturbances in RR cause the value of the RR variable to change, with some variance if we do not act to influence it, and with some other variance if we do act in such a way that we do influence it. If we control the RR variable successfully, the latter variance is smaller than the former.

According to PCT, we control a variable in PR, not RR, and we sometimes compute Quality of Control (QoC) as the ratio of the disturbance-caused variance of the perceptual variable if we don’t act to the variance of the same variable when we act to control it. The question is then about the relation between the RR Quality of Control and the PR QoC. An outside observer cannot determine either kind of QoC, because the observer has perceptual reality different from that of the controller, and has access only to the effect produced by the controller’s actions in RR on the observer’s own unique PR.

An Analyst, however, has no such limitation. By definition, the Analyst has access to all variables related to a problem, in this case, including the values of the RR variable that is responsible for all influences on the sensory system of the organism, the Perception inside the organism, and the context of the perception in the Perceptual World (Perceptual Reality). The PR value of the externalized perception in PR is identical to the perception, and it is the discrepancy between that value and the reference value that generates the error value, and together with its history, the current output value.

The limit to the Quality of Control has two independent components: (1) the variance in how much the RR variable influenced by the control loop’s output changes over the effective loop transport lag time (which includes delays in any integrating or differentiating processes), and (2) the variance in how exact values of the momentary inputs (spikes) influence the exact momentary value of the perception. The second type is what Powers was referring to when he said that he hoped that using “neural current” as a time average over a “bundle” of neurons would not cause more than a 10% error rate.

Only the RR variable affects the sensor values that create the perceptual value, not the externalized perception in its perceptual reality context where it might be, say, the relative locations of a target and a cursor, which might not exist as entities in RR. When one acts to bring the relative location value to a reference, something happens in RR that we perceive as change in that relative location value. Of that we can be assured. Indeed, we can always, at any level of the hierarchy, be assured that if control quality is thus and so, then our actions have caused functional effects in RR that are to that extent correctly modelled in PR. Good control = much knowledge about functional relationships among components of RR.

What we cannot know is what the components of RR might be, nor how they produce their outputs given their inputs. We can, however — and this is the business of much of science — find sets of functional relationships among perceived sub-components that would produce these same functional relationships among the components. The position of a perceiver is closely analogous to that of a programmer trying to analyze an undocumented complicated programme written in an unknown, perhaps Object Oriented Programming (OOP), language.

The programmer could unravel the program into the input-output (functional) relationships among the objects by reprogramming in OOP those same observed functional relationships, and could do it in terms of ever simpler component objects that are re-used in different objects. The programmers version would behave just like the subject programme, but might be programmed entirely differently. The original program might not even have been written using objects in OOP form.

We are in that programmer’s position with regard to what we can (n principle) and cannot (even in principle) know about Real Reality. Our ability to control precisely sets limits on the ways RR interactions might play out in producing the perceptions (replicated in PR) that we control. The Perceptual Functions that produce precisely controllable perceptions necessarily are closely similar to corresponding functional relationships in RR. Our Perceptual Functions have evolved and been tuned and perhaps built from scratch in ways that improve the survival to reproduce of all extant species. What we all can control depends entirely on how well out perceptual functions match Real Reality, and since we are here, that is likely to be “pretty well”.

Martin

1 Like
6

[quote=“MartinT, post:5, topic:15681, full:true”]
[Martin Taylor 2020.10.22.12.20]

MT: What we all can control depends entirely on how well out perceptual functions match Real Reality, and since we are here, that is likely to be “pretty well”.

RM: What does it mean for a perceptual function to match real reality? I don’t recognize the idea that “control depends entirely on how well perceptual functions match Real Reality” as something that can be derived from PCT. From conventional psychophysics, perhaps, but not PCT. I think PCT would say something more like: control depends entirely on perceiving aspects of the environment that can be affected appropriately (in a negative feedback sense) by the outputs of the control system. Controlling perceptions that correspond to what is really out there makes no sense because all that is really out there is what is described by the current models of physics. There are no words, only pressure variations in the air. There are no people, only collections of atoms. It’s all perception, remember.

7

I quite agree. All we are arguing about seems to be what it is that our actions directly influence, which, to my mind, ican only be whatever is really “out there”, not what we perceive. Did you ever hear of a mirage? Or an illusion?

Martin

1 Like
8

Rick, I am happy I was not eating or drinking when I read you last sentences:

“…all that is really out there is what is described by the current models of physics. There are no words, only pressure variations in the air. There are no people, only collections of atoms.”

I would like to see a physicist (a non-mad scientist) who would agree that words and people do not exist. I think that any physicist understands that if “It’s all perception, remember.” then also pressure variations in the air and atoms are perceptions. And they also understand that if there can be pressure variations in the air then there must be some relations between the atoms so that they form different structures. Words and people are based on these structures as much as individual atoms. According to the current models of physics we could calculate how many atoms there are in one person, say x atoms. Now a random lump of x atoms will probably not control its perceptions and be perceived as a human being by any input functions.

If you perceive a human being, there probably is that special kind of structure of atoms in the real reality, not just x random atoms which your input functions happen to pick up. Does that make any sense?

9

MT: The argument I would use is that we perceive and directly control a variable property of something in Perceived Reality (PR), but can act only on whatever is really “out there”, a Real Reality (RR) variable. Disturbances in RR cause the value of the RR variable to change, with some variance if we do not act to influence it, and with some other variance if we do act in such a way that we do influence it. If we control the RR variable successfully, the latter variance is smaller than the former.

Why even use the difference between perceived reality and real reality? There are “perception” and “input quantity” playing the same roles. You can only talk about the real reality using models of the real reality. So, quantities, atoms, molecules, whatever, it’s all models (and all models are perceptions on some level… programs? system concepts?)

To me, it looks like the problem is identifying the controlled variable with the input quantity. The perception is the controlled variable, and then the input quantity could be identical to perception or not, depending on the nature of the input function, how we make the model, etc. One example that shows how qi is not necessarily controlled is using a rate sensor (a differentiator) as the input function.

image
Each box is a function, and each line with an arrow is a variable. Each function corresponds to the following formulas, if we take d() to be differentiation:

Input function (box between qi and p):
p = d(qi) / dt

Comparator (circle up left):
e = r - p

Amplifier (triangle, parameters K for gain and tc for time constant/slowing )
d(v) = (K * e - v) * dt / tc

Integrator in the output function (box between v and qo):
qo = integral(v, dt) or if differentiate both sides:
d(qo) = v * dt

And finally the summing function for the disturbance and output:
qi = qo + qd

In code: python code There you can view the code. To run it, go to File->Save a copy [somewhere]. Then in the upper left click the “play” (>) symbol to run the code.

Here are the results:

image
image767×278 25.7 KB

This could be a person in a car, for example. He is maintaining his speed at 10 units/s, as we can see on the left plot the yellow and blue line are quite near each other, and the error is near zero. On the right plot, we can see that the input quantity is constantly rising, while the output is countering the disturbance (green sine wave). This could mean that the car is acting as an integrator, converting the speed to position.

So, in this case the input quantity, the variable entering the input function is a position, and the perception is the rate of change. They are not even correlated, the input position is not controlled.

However, this is not the only possible model of speed control. We can also represent the input function as identity, so that p = qi; we can loose the integrator from the output function, and then we have all environment quantities representing speed. For that model we could say that the input quantity is the controlled variable, but that is a bit forced by the model of the input function, which simply states that perception is equal to the input quantity.

10

Let me reiterate or make more explicit four things intended by the dyadic diagram.

image001161

  1. It is essential to include the second control system, the observer-experimenter. This is obvious in the agreement that q.i is the observer-experimenter’s perception. In addition, d rather than o is the observer-experimenter’s control output, and d is in conflict with the subject’s control. The diagram should include higher levels. At the relationship level is control of the conflict d, control at the sequence level limits that conflict to just enough to observe the subject’s resistance to d, and control of a principle of science verifies by doing this more than once. Related principles try alternatives and try to falsify the hypothesized CV. This is all implicit but relevant, just as higher levels governing the subject’s control of p are implicit but irrelevant to the immediate test.
  2. There is no claim that the little green rectangle is present in the environment as perceived. In the diagram, it represents the experience of perception by the subject and the experience of perception by the observer-experimenter. We conveniently talk of this as their projection of their perception into the environment or their assumption that their experienced perception is in fact that which they perceive.
  3. To include in the diagram micro-variables that the physical sciences say are present in the environment (air pressures, etc.) we would have to include at least one Scientist control system controlling those perceptions. The absence of this, and its irrelevance to any diagram of the TCV, makes it obvious that in practice we do not take any account of such micro-variables in PCT experiments. We resort to imagining them, on authority of epistemologically prior sciences (physics and chemistry are customarily invoked), only in these philosophical discussions of what is really real and what in the environment is affected by disturbances and control outputs such that a perception is controlled. Let us be clear that these are discussions of the philosophical implications of PCT that have no immediate relevance to doing science.
  4. To say that the TCV is predicated upon the subject and the observer-experimenter having like-structured perceptual input functions would make impossible PCT research on living things other than humans, and denies variation of perceptual input functions across the human population and between individuals. It is essential to the TCV to reconstruct a perception of the subject’s perception from the subject’s point of view. Mere assumption is not enough. This is increasingly the case as we investigate perceptions higher in the hierarchy.
11

RM: As you can see in my diagram, I show actions influencing whatever is really out there (the physical variables x1, x3 and x5). So our argument isn’t about that. It’s about the idea that the “precision with which the perceptual variable’s Perceived Reality corresponds to that of a variable in Real Reality determines the best possible Quality of Control”.

RM: There is no question of a correspondence between perceived and real reality in PCT. We control perceptual variables that are analogs of various aspects of real reality; our perceptual functions define the aspects of real reality that is being controlled. Quality of control – how well we control a particular perceptual variable – depends on the parameters of the control loop that is used to control that variable, parameters that include the physical links between system output and perceptual input.

12

RM: I’m really sorry about that but whatever you were doing other than eating and drinking when you read it seems to have had hallucinatory effects since I never said that words and people don’t exist. I said they don’t exist “out there” in the world described by the current models of physics.

EP: I think that any physicist understands that if “It’s all perception, remember.” then also pressure variations in the air and atoms are perceptions.

RM: Yes. As I said, we know real reality only as a model, which, of course, is a perception. But it is the physics model that model is the environment portion of the PCT model of behavior.

EP: Now a random lump of x atoms will probably not control its perceptions and be perceived as a human being by any input functions.

RM: Yes, a very important point of PCT is that it is not the matter of which a system is made of that matters but how that matter is organized that matters.

EP: If you perceive a human being, there probably is that special kind of structure of atoms in the real reality, not just x random atoms which your input functions happen to pick up. Does that make any sense?

RM: Of course. But that is not really relevant to the point I was making, which is that words and people are perceptual constructions (the outputs of perceptual functions) that are based on an external reality that, from the point of view of physics, is a world of forces, masses, acoustic and electromagnetic energies and the like. That real world certainly provides the possibilities for perceiving words and people. But words and people exist only in systems with perceptual functions that can construct them from the raw material of real reality.

RM: Reality is to organisms like the image on a computer screen is to the computer: a booming, buzzing confusion. From that reality human perceptual systems construct words and people and computer algorithms can now construct them as well.

RM: And, of course, the perceptions that are constructed by organisms – the perceptual variables they control – can’t be arbitrary; organisms must perceive aspects of the world that are “adaptive” in the sense that the ability to control those aspects of the world allows them to survive at least long enough to reproduce. Based largely on introspection, Powers’ hypothesized that the aspects of the world that have proved to be the one’s that are adaptive for human survival are the types of perceptual variables that make up the human control hierarchy (as described in B:CP): intensities, sensations, configurations, transitions, sequences, relationships…etc.

RM: Powers believed that it was perfectly possibly that there are other perfectly adaptive ways to perceive the world that might have evolved instead of the way with which we are familiar. I mention this only to emphasize the fact that there is no concept in PCT of how well we control having anything to do with how well perceived reality corresponds to real reality. How well we control depends on constructing perceptions that are controllable in the context of the constraints of real reality and the parameters of the systems doing the controlling.

Best

Rick

13

RM: I think it is obvious that you don’t need to include the Observer’s entire control system. You only need to show the Observer’s perceptual function, as I do in my Figure:
image

BN: In addition, d rather than o is the observer-experimenter’s control output, and d is in conflict with the subject’s control.

RM: If done properly, the TCV involves no conflict between Observer and Subject. And it is not necessary to show the Observer disturbing the controlled variable in order to see it, anyway. It’s quite possible for the Observer to perceive the variable the Subject is controlling – that is, it is possible for q.i to equal p – when the Observer is in passive observation mode (I’m sure you are familiar with the work of the Plooij’s on using naturally occurring disturbances to see the variables that infant chimps are controlling).

BN: The diagram should include higher levels.

RM: That would be nice but unnecessary to make the point made by my diagram, which is that, when the Observer has identified a controlled variable, q.i, then q.i is equivalent to p and neither correspond to “something in the environment”. q.i and p are the same perceptions because they are the result of equivalent functions of the same variables in the environment (and of the same lower level perceptual variables if they are perceptions above level 1 in the hierarchy).

BN: 2. There is no claim that the little green rectangle is present in the environment as perceived.

RM: I know. But it can give the impression that that is the case, which is why my diagram is much to be preferred.

BN: 3. To include in the diagram micro-variables that the physical sciences say are present in the environment (air pressures, etc.) we would have to include at least one Scientist control system controlling those perceptions. The absence of this, and its irrelevance to any diagram of the TCV, makes it obvious that in practice we do not take any account of such micro-variables in PCT experiments.

RM: I don’t think putting scientists in the diagram would help things. The point of the diagram is to show that q.i and p are equivalent perceptions, in the sense that they are both ultimately the same functions of environmental variables. In practice, Observers doing PCT research do take this implicitly into account by understanding that they have identified a perceptual variable that a Subject is controlling when they “see that the CV [q.i] is protected from disturbances by the actions of the behaving system” [Powers (990331.0033 MST)].

BN: 4. To say that the TCV is predicated upon the subject and the observer-experimenter having like-structured perceptual input functions would make impossible PCT research on living things other than humans, and denies variation of perceptual input functions across the human population and between individuals.

RM: True, which is why I never said that. I said (paraphrasing Bill) that the Observer’s perception of the controlled variable (q.i) will covary with that of the Subject ( p ) if the observer is using a perceptual system closely similar to that in the behaving system. As I mentioned in my post, this could be an artificial perceptual system, like the one used to perceive the ultrasonic echoes controlled by bats when navigating around in the dark.

Best

Rick

14

Rick,

RM: I’m really sorry about that but whatever you were doing other than eating and drinking when you read it seems to have had hallucinatory effects since I never said that words and people don’t exist. I said they don’t exist “out there” in the world described by the current models of physics.

EP: Yes, just that: ‘…they don’t exist “out there” in the world described by the current models of physics.’ This can be read that that they do exist somewhere just like Santa Claus and unicorns, but not in the real physical world. However, the existence of words and people are strictly different from that of Santa Claus and unicorns just in relation to the physical world. If a physician were interested, she had no problems to study and model empirically the physical properties of a human being or a written word – but only as a thought experiment those of Santa Claus or a unicorn.

RM: [snip] But that is not really relevant to the point I was making, which is that words and people are perceptual constructions (the outputs of perceptual functions) that are based on an external reality that, from the point of view of physics, is a world of forces, masses, acoustic and electromagnetic energies and the like. That real world certainly provides the possibilities for perceiving words and people. But words and people exist only in systems with perceptual functions that can construct them from the raw material of real reality.

EP: Yes, but here I must add that a certain part of the real world provides the possibilities for perceiving words and people while another part provides possibilities to for perceiving chords and monkeys.

RM: Reality is to organisms like the image on a computer screen is to the computer: a booming, buzzing confusion. From that reality human perceptual systems construct words and people and computer algorithms can now construct them as well.

EP: And here I must add that (well-functioning) human perceptual systems and computer algorithms can construct words only it there happen to be “words” (= certain kinds of structures) in the image. If there is only a booming, buzzing confusion, then these systems construct only random words or no words at all.

RM: And, of course, the perceptions that are constructed by organisms – the perceptual variables they control – can’t be arbitrary; organisms must perceive aspects of the world that are “adaptive” in the sense that the ability to control those aspects of the world allows them to survive at least long enough to reproduce. Based largely on introspection, Powers’ hypothesized that the aspects of the world that have proved to be the one’s that are adaptive for human survival are the types of perceptual variables that make up the human control hierarchy (as described in B:CP): intensities, sensations, configurations, transitions, sequences, relationships…etc.

EP: What could that “adaptivity” depend on? One preconditions for it is probably that the sensory organs work reliably: that a perception really is some function of the (variance of the) physical stimulus caused by some force in the environment of the sensory organ. That means that there is in this lowest level a correspondence between RR and perception, doesn’t it?

EP: The higher-level perceptions are constructed from the raw material provided by the lower-level perceptions. They are structures made from low level perceptions. Isn’t it quite credible idea that there could be some correspondence (no isomorphy is required) between the structures of higher level perceptions and the structures of the real reality? At least this correspondence would explain the adaptivity.

RM: Powers believed that it was perfectly possibly that there are other perfectly adaptive ways to perceive the world that might have evolved instead of the way with which we are familiar.

EP: Isn’t that a self-evident case? Insects, bats, fishes certainly construct different perceptions, and they are adaptive? But this is partly explained that they perceive and need to perceive partly different structures in the reality.

RM: I mention this only to emphasize the fact that there is no concept in PCT of how well we control having anything to do with how well perceived reality corresponds to real reality. How well we control depends on constructing perceptions that are controllable in the context of the constraints of real reality and the parameters of the systems doing the controlling.

EP: I see here an implicit contradiction because “constructing perceptions that are controllable in the context of the constraints of real reality” depends (partly) on “how well perceived reality corresponds to real reality”.

Eetu

Best

Rick

15

BP: The perceptual signal represents the controlling system’s only knowledge of the controlled quantity. What the system controls, therefore, is the state of the perceptual signal, not necessarily the state of the external (observable to others) controlled variable qi. If the sensor calibration drifts, the perceptual signal will still be maintained in a match with the reference signal, while the visible controlled quantity’s value changes. The variable most reliably controlled by this system is the perceptual signal. Thus the name of my first book: Behavior: the control of perception.

That is pretty confusing, Bill.

Both qi, the input quantity and p, the perceptual signal are called controlled variables, but qi can be controlled or not controlled reliably depending on the calibration of the sensor; and yet it is still called the controlled quantity.

16

Very important observation, Adam! I have been thinking just the same. Powers was a genius but not always very consistent and strict with concepts.

AM: Both qi, the input quantity and p, the perceptual signal are called controlled variables, but qi can be controlled or not controlled reliably depending on the calibration of the sensor; and yet it is still called the controlled quantity.

EP: I would like to tease the tangle as follows:

qi, the input quantity is the value of an input variable and it means the amount (or strength) of the environmental stimulus to the sensory organ.

Input function transforms this value to the value (strength) of the perceptual signal. This result of the transformation, the value of perceptual signal, is the perception which is some (mathematical) function of the input quantity.

The latter, the perception is p-controlled (see my message [Eetu Pikkarainen 2020-10-26_13:08:51 UTC] in CSGnet) and the previous, the input quantity is e-controlled. E-control is not real control but rather like stabilization (to some value) or something like that.

EP: That quotation nicely points that Rick’s claim that “there is no concept in PCT of how well we control having anything to do with how well perceived reality corresponds to real reality.” is at least questionable.

17

The parameters of control systems are among “the constraints of real reality”. So you are making a distinction between “correspondence” of perceptions to real reality and “controllability” of perceptions in respect to real reality. What difference does this verbal distinction make?

From a denial that they are synonymous it follows that control of perceptions tells us nothing about real reality. In other words, it says there is no point to doing science. We’re just making it all up, all the theories and models of science are fabrications of the imagination with no valid claim to correspond to reality.

Accepting that they are synonymous–that controllability within reality demonstrates correspondence to reality–does not go to the opposite extreme. The degree of correspondence to reality is greater or lesser according to the degree of controllability. And that is the nature of science.

18

One of the first things we check in the search for the controlled variable is whether the organism can perceive the posited CV as we perceive it. For example, does blocking that sense modality interfere with control? If the response profile of the sensor that we have identified is disproportionate to the values of the q.i variable that we have identified, but control is not correspondingly poor, does some other sensory modality provide input to the perceptual input function for p? For example, the CV for speech perception depends simultaneously on visual as well as auditory input (cp. the McGurk effect) so q.i for perception of a phonemic contrast is not a single simple quantity.

19

Sorry, I must still continue. In that quotation Powers says that “What the system controls, therefore, is the state of the perceptual signal, not necessarily the state of the external (observable to others) controlled variable qi.” So at least here qi is not the observer’s perception but an external variable which is observable to others. Observer’s perception is not observable to others.

Inspired by your images (RM, BN, and especially AM) I made that below:

image003.png

Above the upper blue line is the controller and below the lower one is the observer. The upper left box is controller’s input function which creates the controller’s perception from the controller’s input quantity (qic). Qi is the value of the (complex) input variable (consisting of values of x, y and z depicted as green arrows) – or the corresponding environmental variable CEV. Both controller and observer have roughly same qi depending on the similarity of their input functions and their relationships to X. Right side box is controller’s output function which creates the output quantity form the error. The output quantity can divide to different effects partly affecting the X and partly something else and being thus side-effects.

Now what is that X? It is an unknown environmental structure which mediates effects and thus creates the feedback function. It can be as complex or as simple as ever, but it determines how the output effects affect the input quantity and thus the quality and the whole possibility of control. Presumably it consists basically of atoms – or rather something which physics models as atoms. It can then contain molecules, liquids, cells, organs, organism, machines etc. We cannot know what its ultimate character is, but we can learn to perceive how it works and how it can be controlled.

Successful control requires that we perceive X well enough and affect it in right manner and strength. We can make mistakes in both sides of control: input and output. We can perceive wrongly or affect wrongly (and if we perceive wrongly then we quite surely also affect wrongly), but from mistakes we can learn and have learned. Take for example politics. If something disturbs us: taxation, injustice, or any we must know what the system is kind we want to affect. We can bite the tooth or hit our neighbor, but it doesn’t help much.

20

RM: What’s the context of the quote? That might make it less confusing.

21

Sorry, here is the source of the quote. Looks like Bill was writing a short description of PCT.

22

It might be worth noting that whatever qi might be for any particular controlled perception, it is not located at the sensory periphery unless you are talking about controlling the output of a single visual rod or cone, or auditory hair cell or the equivalent in the other senses. Even then, the output of that rod or hair cell is not a function of its present input. It depends on the time since the last firing, on firings of neighbouring sensors, which we can lump together as “spatial-temporal context”.

That context strongly affects the relationship between input and output at any place on the sensory surface that contributes to any perception that is controllable through the environment (i.e. any perception that is not entirely imaginary or hallucinatory), if the environment is taken to include internal sensations as well as sensations from outside the “skin bag” (meaning that the alimentary tract is takien to be “inside” the skin bag but “outside” the perceptual function, comparator, output function triad that forms an “Elementary Control Unit”. Only at what Powers called the “controlled variable” (CV) and I call the Controlled Environmental Variable (CEV) do we find the convergence to a scalar value of a property of something in Real Reality that we can legitimately call qi…

One must locate qi at the environmental variable that causes the perception being controlled, as the value of a property of some entity. Just as the perceptual value is a function of many sensory inputs and inputs from imagination, so also is the effect on qi a function (determined by the flow of forces and chemical influences through the environmental feedback path of the many material, muscular, and chemical outputs from the body that houses the control unit in question. Adam’s diagram hints at this, but I would suggest one edit in it. There is no guarantee that the function created by the observer/experimenter’s input processing uses all the same lower-level variables as does the controller in question, so I would either delete one arc to the observer or add another one.

The point here is that no matter what constellation of influences converge to create qo at the CEV (the effect on some property of something in the environment), the resulting qi goes through many different sensory systems before reaching the hierarchy of percpetual functions that converge to create what we perceive and project into perceptual reality, and those diverging and converging pathways (braided, to use Warren McCullough’s terminology) vary significantly over a wide range of time scales.

Over evolutionary time, our systems have been built to produce consistent effects from what on the surface looks like a big mess, and to produce perceptions (at least at lower levels) that functionally correspond very closely to the functionality of the real world. The rest we build by reorganization within an individual lifetime. The wonder is that it has taken only a couple of billion years to achieve this level of operational stability, and that qi does usually seem to correspond pretty closely with p (at low levels of perception).

23

Eetu, interesting diagrams
Some more input, this is from A Primer For Programmers, draft, BP, emphasis by me,

BP: Mon Nov 09, 1992
Suggestions for standard terminology

A function is a physical device with an output signal the magnitude of which can be
computed from the state of its input magnitudes. All functions are true mathematical
functions: that is, they may have multiple inputs (arguments) but they produce only one
output (value of the function given those arguments). Thus the term function refers both
to some physical element of the system and to the equivalent mathematical function that
describes the dependence of its output on its input(s) in terms of magnitudes.

A generic control system consists of an input function, a comparator, and an output
function. The output of one function generates a variable that is an input to another
function. Such information-carrying variables inside the system are called signals. A
signal not only represents the value of the function, but serves to carry that value to
the input of another function in a different physical location. All signals have a single
measure, magnitude. The name of a signal identifies a pathway; the value of the signal
indicates the momentary magnitude of the signal carried unidirectionally by that pathway.

The environment model

In the environment of a control system the variables are called quantities. The output of
the output function is measured in terms of an effect on a physical variable called the
output quantity. The output function in a model of a single control system interacting
with an environment is therefore a transducer: its input is a signal while its output is
a quantity. The output quantity is always defined so that its magnitude depends only on
the output function’s value: it is always a single variable. If it has multiple effects
in the environment, each of those effects must be separately indicated in a model of the
environment.

The input to the control system is another physical variable called the input quantity.
The input function senses the state of the input quantity and converts it to a perceptual
signal. The input function is also a transducer in a single system-environment model; its
input is a physical quantity and its output is a signal. An input function may respond to
multiple input quantities.

In the environment, there is a feedback link connecting the output quantity to the input
quantity. This link is called the environmental feedback function, or simply the feedback
function.

Also in the environment there is a link through which independent environmental variables
called disturbing quantities act on the input quantity concurrently with the action of
the output quantity on the input quantity. Because the number and kind of disturbing
quantities is immaterial, it is customary, when modeling a single control system, to
represent all disturbing quantities and their individual links to the input quantity
as a single equivalent disturbance acting through a single equivalent disturbing function.

The control system model

The perceptual signal generated by the input quantity enters a comparator; also entering
the comparator is a reference signal, an independent variable. Where possible, the signs
of various system constants are chosen so that the reference signal has a positive effect
on the comparator while the perceptual signal has a negative effect. The comparator is a
function with two arguments and a single value. The output value is represented by an
error signal, the magnitude of which is equal to the reference signal’s magnitude minus
the perceptual signal’s magnitude. The error signal enters the output function. Often, as
shorthand, we speak of subtracting one signal from another, or adding signals together.
What is meant is that the magnitudes are subtracted or added.

The system-environment diagram

                           sr 
                 sp        +|      se 
              -----------> (fc)-----------> 
             |           -                 |    CONTROL 
            (fi)                          (fo)   SYSTEM 
-----------  | --------------------------- | ------------- 
            qi <-----------(ff)<--------- qo   ENVIRONMENT 
             ^ 
             | 
            (fd) 
             ^ 
             | 
            qd

DEFINITIONS:

Signals:
sp = perceptual signal
sr = reference signal
se = error signal

Functions:
fi = input function
fc = comparison function or comparator
fo = output function
ff = feedback function
fd = disturbance function

Quantities:
qi = input quantity
qo = output quantity
qd = disturbing quantity

THE CONTROL EQUATIONS:

System:
sp = fi(qi)
se = sr - sp

Interface transducers:
qo = fo(se)
sp = fi(qi)

Environment:
qi = ff(qo) + fd(qd)

Combined equations:
System: qo = fo(sr - fi(qi))
Env: qi = ff(qo) + fd(qd)

AM: I made some changes to the diagram, to make the adding function (fa) in the environment explicitly drawn, and varibles qi, qd and qf written next to “lines”, following the same convention as inside the control system. If the comparator gets a special box, all it does is subtract the sp from sr, so should an adder, as it adds qf and d. All the equations defined by Bill previously still hold, this is just a diagram change. Maybe I’d be happier if names qd and d would switch, though.

                           sr 
                 sp        +|      se 
              ----------->(fc )-----------> 
             |           -                 |    CONTROL 
            (fi)                          (fo)   SYSTEM 
-----------  | --------------------------- | ------------- 
          qi |                             |    ENVIRONMENT 
             |    +                        |
            (fa) <--------(ff)<-------------  
             ^ +     qf              qo
           d | 
            (fd) 
             ^ 
          qd |

AM: One of the interesting points Bill made is that there could be multiple input quantities entering the input function. We could call them x1, x2, x2, or we could call them qi1, qi2, qi3. If the perceptual signal p or sp is a function of those input quantities, then how do we call the hypothetical controlled variable in the Experimenter, that is a function of the same input quantities? I guess “controlled quantity”, or qc is ok? The general process of TCV is still the same, it is just switching up the names a bit.

24

RM: Yes, this quotation of Bill’s seems to exactly contradict his statement I posted: “The CV” is the observer’s perception." But later in the post where Bill says what you like – that the state of the perceptual signal is not necessarily the state of the external (observable to others) controlled variable qi – he says this:

BP: The Test requires applying disturbances to the supposed controlled variable, and seeing whether the system’s output action varies so as to have an equal and opposite effect on it. When such a relationship is found, we presume that there is a perception inside the system corresponding to the observable variable being controlled relative to an internal reference signal.

RM: In other words, when we have successfully done the Test we presume that qi (the observable variable being controlled) corresponds to p (the perception inside the system). So that’s two statements saying that qi = p to one saying qi<>p.

RM: My guess is that when Bill says:

BP: “What the system controls, therefore, is the state of the perceptual signal, not necessarily the state of the external (observable to others) controlled variable qi.”

he is referring to how the magnitude of p relates to the magnitude of qi. He is not referring to how the magnitude of qi relates to the type of variable represented by qi. The evidence for this is when Bill says:

BP: If the sensor calibration drifts, the perceptual signal will still be maintained in a match with the reference signal, while the visible controlled quantity’s value changes.

RM: Sensor calibration is not the same as the perceptual functions that define the controlled variables, qi. Sensor calibration (in PCT) defines the quantitative relationship between neural firing rate ( p) and the input to the sensor, qi: in the simplest case p = k*qi. Perceptual functions, on the other hand, define the qualitative nature of qi. For example, a perceptual function that computes t - c defines qi as a perception of the distance between target and cursor.

RM: When calibration drifts, there is a change in k, which changes the quantitative relationship between neural firing rate ( p) and controlled variable (qi). What Bill is saying is that if k drifts lower, for example, then p will remain matching the reference signal but it will do so because the control system will have changed qi (made it large). If qi is t-c then the downward drift of k will produce an increase in the distance between t and c. But this is just a change in the magnitude, not in the nature, of qi. The observer still sees that the system is controlling a perception of t - c; the calibration change results only ina change in the apparent reference state of this perception.

RM: So I see no inconsistency between Bill’s two statements: The one I posted where he said that qi = p and the one Adam posted where he said that qi<>p. There is no inconsistency because qi = p is true qualitatively and qi <> p is true quantitatively; qi = p because qi is an analog of the same aspect of the environment as p; qi<>p to the extent that there is a difference in calibration of the perceptual systems of observer and control system.

EP: Successful control requires that we perceive X well enough and affect it in right manner and strength.

RM: How can you know how accurately a person is perceiving X (in the qualitative sense) when, by your own arguments here, that perception may itself be inaccurate?

Best

Rick