* Diagram Police *

[rsmarken]

@EetuP

Rick,

it seems that you are quite firmly stuck to epistemological idealism: real environment is not needed or at least we should not talk about it.

Well, I’ve been called a lot of things in this group – bullshitter, persistent mistake maker, clown, etc – but “epistemological idealist” is just going too far

I actually do think there is a real, physical environment on the other side of our senses. But I agree that my way of talking about perception may have made it difficult to see that that is the case. So I’ll try to explain my point of view – which, I’m pretty sure is the PCT point of view – using your nice diagram.

@EetuP

Here is the way how I could try to understand your way of thinking and talking in a somewhat consistent way (as for qi):

The question marks are there just to show the variables which have no name and about which in PCT should be talked about. What is not consistent is that there are two q variables in the environment (and as such they should be hidden) and one in the observer.

This diagram is actually a pretty good description of my point of view. Inside the open circle in the environment that represents the controlled variable is where the world of physical variables would reside; these are the environmental variables on which the perception of the controlled variable is based.

In the tracking task, these physical variables would be the spatial distribution of light emitted from the computer screen. In Powers’ (1973) Science paper those variables are labeled as v’s:

In this diagram, the “Input quantity” is equivalent to what we have been calling controlled variable, q.i (the q we use for labeling variables is a legacy of Bill’s use of “quantity” to refer to variables in the environment).

Because the other relevant environmental variables in the diagram – output and disturbance – are labeled q.o and q.d (which is qe in your diagram), it’s natural to think of q.i (the “Input quantity” circle in the diagram above) as also being a variable in the environment. But it is not! This can be seen from the arrows leading from the physical variables inside the Input quantity circle, the v’s, to the Sensor function (now called the Input function). The Sensor signal coming out of the Sensor function is what we now call the perceptual signal and the diagram shows that this signal is constructed by the Sensor function, f, from the sensory effects of the physical variables, v1, v2,…vn:

So the circle around the v’s in Bill’s diagram represents the perceptual variable that is constructed by the Sensory (Input) function, f (x1,x2…xn). It is called an Input quantity (now contorlled variable) and it is show in the environment but it is not something that exists as an entity in the environment.
Rather, it is a function of physical variables, v1, v2…vn, in the environment. And this function – a perceptual function – is carried out by the control system; and it can also be carried out by an observer of the control system who possesses (or can construct) a Sensory function equivalent to the one used by the control system.

So with only slight modification of your diagram – adding v’s inside the circle that appears in the environment and replacing the question marks with arrows running from the v’s in the circle to the Input functions of the controller and observer – you will have a correct representation of my view of the relationship between people (controllers and the observers thereof) and the environment.

@EetuP

And if, as you say, qi is “controlled variable” then what the controller controls is an other person’s perceptions, which sounds really mystical!

Actually, that problem doesn’t exists because q.i (the controlled variable or Input variable in the diagrams) is actually a perceptual function of the sensory effects of variables in the physical world --the v’s in Bill’s diagram – the same physical world that is the basis of the perception of both controller and observer. Of course, observers (PCT researchers) have to take into account the fact that they are typically perceiving that world from a different perspective than the controllers they observe. But this is usually pretty easy to figure out.

I think what may be the reason you suspect me of being an epistemological idealist is that I rarely describe controlled variables in terms of the physical variables of which they are a function. I think once we get above the level of what Bill called “intensity” and “sensation” type perceptions it just becomes too tedious to do that. So we talk about “target lines” and “cursor lines” (rather than rows of low illumination levels) as being the basis of a perception of distance. Of course, the target and cursor lines are themselves perceptions.

I hope this makes the PCT epistemology a bit clearer but, if not, I hope it at least shows that, while I may be many things, I am not an epistemological idealist.

Best regards, Rick

[wmansell]

Thanks Rick, really helpful.

[EetuP]

Rick,

as you say: “but “epistemological idealist” is just going too far” and I am sorry for that. As I replied to Warren, the problem is how we talk. Because PCT is in a way a difficult area for a colloquial mind we should be very careful to keep our talk as well as diagrams consistent.

Very good starting point is Adam’s “The diagram is marked from the perspective of the researcher, and the system/environment line is a limit between environmental variables and neural variables.” earlier in this thread. So the variables like p, r, and e are neural variables, and qo, qe (sometimes qd), and qi are environmental variables.

[EetuP]

[For some reason Discourse cut away the end of my email reply, so it is here:]

The problem you take up that qi is a complex variable (consisting of multiple v’s in that Powers’s diagram) is common to all environmental variables (and in addition to neural variables too). It gives no reason to treat qi differently. As well as “neural current” is a convenient way to talk about a (fuzzy) bundle of parallelly spiking neurons, the environmental variables are or at least often can be “environmental currents” of parallelly affecting effects. Even in a simple tracking task the output (variable) consists of an interplay of many muscles, and the disturbance consists of an interplay of many electrical currents in the computer, and similarly the input variable consists of an interplay of many light beams in addition to other possible effects to the sense organs of the subject. There is no reason to treat qi differently from the other environmental variables.

The circle in my diagram is (like they always should be) an adder function, which collects together all the needed environmental sub-variables (coming from outputs and disturbances). Needed for what? This is the tricky point. Needed for the input function. This is the way how the input function defines qi: by selecting the environmental sub-variables and then constructing the p from them.

So qi (as an environmental basis for p) is in an environment. Note that if for example those variables t and c in your tracking example are in the environment then there are also functions (t-c), (t+c), (t*c), (t/c) ect. in the environment as complex variables, too! It is just that the input function (as an organ) takes one or some of these complex environmental variables and defines them as qi or Corresponding Environmental Variable (CEV) AND produces an analogical neural current which is marked as p.

Eetu

[rsmarken]

@EetuP

So qi (as an environmental basis for p) is in an environment. Note that if for example those variables t and c in your tracking example are in the environment then there are also functions (t-c), (t+c), (t*c), (t/c) ect. in the environment as complex variables, too!

It seems strange to me to say that functions of environmental variables, such as (t-c), (t+c), (t*c), (t/c), etc., are in the environment. Those are what I would call perceptions. But the fact that you recognize that there are different possible functions of the same environmental variables is a very promising development.

@EetuP

It is just that the input function (as an organ) takes one or some of these complex environmental variables and defines them as qi or Corresponding Environmental Variable (CEV) AND produces an analogical neural current which is marked as p.

So your view is that input functions are detectors rather than constructors. The only problem with this for me is that it seems to point the study of living control systems away from what I think is the main question about their behavior that is posed by PCT: what perceptual variables do organisms control when we see them carrying out various behaviors?

Powers hypothesized that organisms (humans in particular) come prewired to construct 9 or 10 different types of perceptual variable. The main goal of PCT research is to test this hypothesis.

Seeing perception as a process of detection, on the other hand, implies that you already know what variables are “out there” to be detected and your research will be aimed at determining how well organisms can detect those variables. That is, your research will be a lot like conventional psychophysics, possibly with the added twist of seeing how the organism’s ability to detect those variables affects their ability to control them.

So I think this apparently rather academic difference between seeing perception as a process of detection versus seeing it as a process of construction has very significant implications for how we do the science of PCT. I guess what I am stuck in is the notion that perception is a process of construction rather than detection. And this explains why I have such a bad reaction to the concept of a CEV, which implies that the variables we control are “out there” (CEV’s) to be detected by the controlling system; I think the variables we control are constructed “in here” from the sensory effects of environment variables that are “out there” and the job of PCT science (using the TCV) is to find out what these controlled variables are.

Best, Rick

[EetuP]

1. Constructivism, in a radical form, represents epistemological idealism. It should not be mixed with ontological idealism or solipsism, which assume that nothing else exists than the knowing subject itself. Instead epistemological idealism can admit that there may exist whatever in the environment of the knowing subject, but the knowledge of the subject does not depend or is not determined by the environment – the knowledge is based totally on the properties of the subject itself. These all are respectable philosophical stances because we have no absolute knockdown arguments against them, but especially constructivism is interesting stance because it needs not be absolutely idealistic but it can admit some – more or less – role to the environment in the production or validity of knowledge. The interesting question is what does “construction” mean?

2. It is an open – at least for me – question of the philosophy of mathematics what functions and how there exist in the environment and what does not. Let’s take a simple example: There are apples in table so that some of them are in left side of the table and the rest are in the right side. Here we have two variables: x (amount of apples in the left side) and y (amount of apples in the right side). But if we have these two, we certainly have also a third, say z = (x+y) as all the apples in the table. As well we have the difference of the amounts w = (|x-y|). Also there are all the apples which are not on the table, apples in northern side of the table etc. etc. endlessly. So every time when we perceive something, even very simple and low level perceptions, like intensities, our perceptual functions select (or choose) some effects from the endless amount of different possibilities in the environment.

3. In B:CP Powers defines: “PERCEPTION: A perceptual signal (inside a system) that is a continuous analog of a state of affairs outside the system.” In the diagrams p means perceptual signal inside a system (either controlling or observing system). qi means that “state of affairs outside the system”. If we assume (note, that it is a theoretical assumption based on perceptions) that p “is a continuous analog of” qi, then qi must be one variable (which can be a function of many other variables) similarly as p is one variable (even though it is a function of properties of many individual neurons). You cannot put qi inside the observer because the controller is NOT (normally) perceiving observer’s neural currents but some state of affairs outside. Observer’s perception (po) of course may be a continuous analog of controllers perception, but the explanation of this is that they both are analogs of the same state of affairs (qi) in their common external environment.

4. There is a decades old debate in CSG mailing list about do we control our perceptions or their environmental correlates or both. Lets not open it again because I think it found a satisfactory solution formulated by Fred if remember right: “We control our perceptions by controlling* our environment.” Only remaining problem is whether that latter “control” marked by an asterisk should be replaced by some other term, like stabilize, affect or something else. Anyway, that necessary connection between perceptions and environment (connected to the idea of TCV) offers the best possible counter argument against different forms of idealism.

5. So what does it mean to “construct” a perception? I regard this as a complex question without one definitive answer. Different perceptions are constructed differently. What there must be always is that selection or choosing from the endless amount of environmental effects and their combinations. This is one stage in the process or phenomenon of how perceptual function defines qi. In addition these effects are weighted in different ways. As these functions are often described there are multipliers and added constants, but certainly much more: any kind of manipulation which neural systems can do. But whatever that manipulation and molding contains, still the analog relationship between qi and p should be maintained.

Here should the Powers hypotheses of hierarchical levels of perceptions be taken up: The levels define different ways to construct. The higher level perceptual functions select some of the lower level perceptions and combine and weight them in a new way. In addition it seems clear that most highest levels do not combine only current low-level perceptions but also memorized earlier perceptions and thus even the analog relationship between these perceptions and the current state of affairs outside the system become questionable.

That is how I think at the moment. But the bottom line is that qi should be situated in the environment in diagrams, and the real or primary controlled variable is p, and there is assumed an analog relationship between qi and both controller’s and observer’s perceptions, and this relationship makes it possible for the observer / tester to determine what the perception of the controller is. (BTW, the measurement of the current value of qi is of course based on observers perception.)

[amatic]

@EetuP

But the bottom line is that qi should be situated in the environment in diagrams, and the real or primary controlled variable is p,

I agree. Check out this one:

Bravely placing qi inside the person. This is figure 3 from the same paper.

I don’t know, it is skipping on the whole perception, error, reference signal thing, I kind of like the simplicity, and I like the use of the adder/comparator for qi and qi*. The definition of qi needs work, it does not make much sense like this with that dash - minus after the definition psi(qe).

Maybe the whole person box is unnecessary, so everything is in the ‘environment’. I bet the stumbling block was where to put the reference level. The reference signal r, a neural quantity, is always inside the organism, as the cannon goes, but where the hell does the reference level go?

Another option is to put multiple perceptual variables inside the person box. Not sure if that is a better solution, but the paper does state that PSI(qe) and PSI(qo) are psychological magnitudes.

The point of Figure 3 is that the stimulus-response function is not the organism function, but the inverse of the environmental function. It is not clear what are the environmental functions here, because the PSI1 and PSI2 are defined as:
PSI1(qe) = k1 * log(qe)
PSI2(qo) = k2 * log(qo)

According to both figure 3, my second diagram, and the text of the paper, these functions are inside the organism, as perceptual input functions. So… the qo-qe relationship reflects the inverse of two perceptual functions, which are properties of the organism, making the case that the Stevens power law is not a behavioral illusion.

Someone needs to perform an experiment with this setup to verify that the proposed relationship really happens, with explicitly identified functions and quantities.

[rsmarken]

@amatic

@EetuP

But the bottom line is that qi should be situated in the environment in diagrams, and the real or primary controlled variable is p,

I agree. Check out this one:

Bravely placing qi inside the person. This is figure 3 from the same paper.

Why all the meanness? Why not just say that I incorrectly placed qi inside the organism? But I think qi is placed correctly in this diagram (and Bill thought so too since we wrote the paper describing this work together) since the controlled variable is hypothesized to be the difference between two perceptual magnitudes – the logs of the stimulus (qe) and response (qo).

@amatic

The point of Figure 3 is that the stimulus-response function is not the organism function, but the inverse of the environmental function.

No, the point is that a the observed S-R power function is the inverse of the feedback function, which in this case includes the perceptual function that takes the log of the numerical magnitude estimate.

@amatic

According to both figure 3, my second diagram, and the text of the paper, these functions are inside the organism, as perceptual input functions. So… the qo-qe relationship reflects the inverse of two perceptual functions, which are properties of the organism, making the case that the Stevens power law is not a behavioral illusion.

Steven’s Power Law is, indeed, a demonstration of a behavioral illusion. The illusion is that the observed power law relationship between qe and qo – known as Steven’s Law – is the actual psychophysical function (which is the organism function under study) when, in fact, the actual organism function is a logarithmic function; the Weber-Fechner Law.

@amatic

Someone needs to perform an experiment with this setup to verify that the proposed relationship really happens, with explicitly identified functions and quantities.

As I said in the paper, it has already been shown (1) that the observed coefficient of Stevens power law can be predicted from the coefficients of the Fechner log function for the variable estimated (qe) and the variable used to make the estimate (qo). There is a considerable research literature on cross-modality matching which shows that this is empirically true.

(1) MacKay, D. M. (1963). Psychophysics of perceived intensity: A theoretical basis for Fechner’s and Stevens’ laws. Science, 139(3560), 1213–1216.

Best, Rick

[EetuP]

@amatic

What is this paper and is it available somewhere?

Eetu

[amatic]

Yeah, that’s Marken, Kennaway and Tauseef 2022. https://journals.sagepub.com/eprint/9TAWRAGJGPA2KJ3UZIVT/full

[EetuP]

Thanks for the reference. This poses interesting questions.

Here is the Marken &al diagram once again:

If I have understood nearly right, then the upper part (a) should describe the original Steven’s Power law (from 1957), that there is an organism function (psi(q_e)= q_e^a) which explains the relationship between q_e (heard sound magnitude) and q_o (subject’s numeric estimation of the heard magnitude). The found relationship was that q_o=q_e^a.

And the lower part (b) describes the result of Mackay’s analysis (from 1963) that rather than one stimulus perception and one response measure there are two perceptions which are compared and their difference controlled in the Steven’s setting. (Mackay does neither mention “control” nor Powers.) The power relationship between q_o and q_e is explained by the relationship (remainder) of the two logarithmic perceptual functions which according to Fechner-Weber law logarithmically create the perceptions (heard sound and seen number).

What I don’t understand is how (b) show’s that (a) is a behavioral illusion (BI)? BI means that someone thinks that a function inside an organism causes (and explains) the relationship between input and output while in fact the relationship is caused by a function in the environment, right? But the both psi functions and also their difference are inside the organism. So, as Adam said, the Mackay’ result shows that Steven’s Power law is NOT a case of BI. (Instead it can be some other kind of illusion, which mixes for example input and output functions.) No environmental feedback function explains the power relationship.

But I do not agree with Adam that qi is place inside the system. There is no qi (literally) place in that diagram. Instead both qe and qo extend inside the systemin and there are no environmental functions. (It seems that Rick, quite inconsistently, thinks that the psi input functions were feedback functions inside the system.) If the experiment situation is thought so that the system controls the difference between the sound perception and visual (number) perception with a zero reference – as if in a tracking task – then there should normally be at least an adder function for qe and qo which produces the difference perception as qi outside the system, like this:

In the experiment the subject first hears a sound and (then?) sees a number which is said to express or correspond the loudness / strength of the sound. Then she hears a newr sound and is asked to show a number which corresponds the strength of that new sound. For me that setting feels somewhat different from tracking tasks. It feels like the subject is not so much controlling the difference between the sound and the number but rather their sameness so that the perception of the sound strength becomes a reference for the perception of the numerical “strength”. Something like this:

Of course the both input functions should be placed there too, and the whole setting should be tested empirically,

but what do you think about that idea?

[rsmarken]

@EetuP

Thanks for the reference. This poses interesting questions.

Here is the Marken &al diagram once again:

If I have understood nearly right, then the upper part (a) should describe the original Steven’s Power law (from 1957), that there is an organism function (psi(q_e)= q_e^a) which explains the relationship between q_e (heard sound magnitude) and q_o (subject’s numeric estimation of the heard magnitude). The found relationship was that q_o=q_e^a.

Yes, you understand it correctly. Stevens Power law, q_o=q_e^a, is taken as a description of the organism function that relates the physical variable, q_e, to the behavioral variable, q_o.

@EetuP

And the lower part (b) describes the result of Mackay’s analysis (from 1963)…

No, that’s our PCT model of what is going on in the magnitude estimation experiment that is used to find the Power Law. Mackay is cited by us because he came to a similar conclusion as we do, but based on somewhat different, non-control theory assumptions.

@EetuP

What I don’t understand is how (b) show’s that (a) is a behavioral illusion (BI)?

The Figure doesn’t show it very well, I’m afraid. We had to trim the paper down considerably, which, I think, made it overall a better paper but we may have trimmed out a bit too much in some places. In this case you have to get the description of the behavioral illusion from one sentence in the text, where it says “equation (6) is the inverse of the logarithmic feedback function that relates qo to qi.” Equation 6 is the power function equation for q_o in the (b) part of the Figure. The power function is the inverse of the logarithmic feedback function.

@EetuP

BI means that someone thinks that a function inside an organism causes (and explains) the relationship between input and output while in fact the relationship is caused by a function in the environment, right?

The BI (behavioral illusion) is taking an observed relationship between environmental and behavioral variables (in this case, between q_e and q_o) for the organism function when, in fact, it is the inverse of the feedback function connecting q_o to q_i. We usually take the feedback function to be equivalent to the environmental connection between q_o and q_i. But this assumes that the physiological connection from the sensory effects of the environmental basis of q_i to its perceptual analog, p, is irrelevant.

In this case this connection is relevant because the presumed logarithmic relationship between the components of q_i – Psi (q_e) and Psi (q_o) – are what produce the observed non-linear (power) relationship between q_e and q_o.

@EetuP

But the both psi functions and also their difference are inside the organism. So, as Adam said, the Mackay’ result shows that Steven’s Power law is NOT a case of BI. (Instead it can be some other kind of illusion, which mixes for example input and output functions.) No environmental feedback function explains the power relationship.

I hope what I said above clarifies things a bit. It’s the feedback function – the function that relates system output to controlled variable – that is the basis of the behavioral illusion. In this case, the fact that the controlled variable is a perceptual variable is important because it is the logarithmic perceptual functions that are in the path of the feedback function connecting q_o back to q_i that produce the illusion.

@EetuP

But I do not agree with Adam that qi is place inside the system. There is no qi (literally) place in that diagram.

As shown in the diagram, q_i is a difference in psychological magnitudes, Psi (q_e) - Psi (q_o). So in this diagram – as in all PCT diagrams – q_i is equivalent to what we usually call p, and is inside the system, as it always is.

@EetuP

Instead both qe and qo extend inside the systemin and there are no environmental functions. (It seems that Rick, quite inconsistently, thinks that the psi input functions were feedback functions inside the system.)

It is unusual because, as I noted above, we usually assume that any non-linearities in the components of feedback functions that transform the sensory effects of environmental variables into their psychological correlates are irrelevant. In this case, those non-linearities (the log functions) are definitely not irrelevant.

@EetuP

In the experiment the subject first hears a sound and (then?) sees a number which is said to express or correspond the loudness / strength of the sound.

This is a good point. A more accurate model of magnitude estimation than the one in the Figure in our paper would take this into account. But the temporal effects that may exist in this task are apparently very small relative to the main result which is a power relationship between q_e and q_o when, in fact, the actual organism function relating q_e to q_o is logarithmic.

Best regards, Rick

[MartinT]

So far as I can see, this thread about diagrams has not made clear (though Eetu tried) the distinction between what happens in the unknown Real Reality between the final, say muscular, effects on the unknown Real environment and the first affected sensors (e.g. rods, cones, and the equivalents in the other senses).

Whatever those sensors produce eventually makes its way through internal (e.g. neural) processes to be translated by an internal perceptual reality (PR), which over time, evolution, and reorganization, becomes controllable to some always imperfect degree.

PR is constructed by our perceptual functions, as Rick has been telling us for many years. But Real Reality is not. It just is what it is, and we shall never know what it is. But we can discover what it does to some level of precision (an incredible level in modern physics).

Those functions in RR are what the perceptual apparatus reorganizes to accommodate, allowing ever better Quality of Control (QoC), the more opportunities it has to develop from experience. Quality of Control, however you measure it, can be only as good as are the relationships between what happens in PR and what really happens in Real Reality as a function of whatever it does to the influences of our control actions on it.

When I want to draw a diagram like those cited in this thread, I always try to show (a) that the input to the organism is an output from RR, which the perceptual system converts to a perception, and (b) the perceptual signal defines the content of the relevant portion of PR.

The closeness of the correlation between the controlled variable P and the corresponding variable in Real Reality determines the limit of QoC (I use “correlation” colloquially, rather than mathematically. I’d prefer to use information-theoretic terms, but that always produces an unnecessary cyclic argument, so I don’t).

[rsmarken]

@MartinT

The closeness of the correlation between the controlled variable P and the corresponding variable in Real Reality determines the limit of QoC…

When you observe a poor QoC, how do you know that the observed QoC is a result of a low correlation between P and the corresponding variable in Real Reality rather than poor (low gain) control of a P that correlates perfectly with the corresponding variable in Real Reality?

[MartinT]

@rsmarken

When you observe a poor QoC, how do you know that the observed QoC is a result of a low correlation between P and the corresponding variable in Real Reality rather than poor (low gain) control of a P that correlates perfectly with the corresponding variable in Real Reality?

You don’t.

[EetuP]

Originally this thread is about drawing consistent diagrams, but because by these diagrams we try to model processes which are embedded in the unknown Real Reality we cannot avoid certain kind of “metaphysical” questions.

As for Rick’s last question, I agree with Martin that we cannot know accurately and for sure “that the observed QoC is a result of a low correlation between P and the corresponding variable in Real Reality rather than poor (low gain) control of a P that correlates perfectly with the corresponding variable in Real Reality” but perhaps even that can be vaguely modelled by testing: If rising the gain (development of output) increases the QoC then the low QoC was more a result of low gain but if that has no effect and if yet corrections in the input function increases the QoC then probably the correlation was low. Actually the possibility of control depends always on two correlations (still in a colloquial sense): the perception must correlate with Real Reality and also the output must correlate with it. The QoC depends on these both.

I must apologize that when writing my last message I was blind and did not see that there really was literally the qi written inside the system in the Rick’s diagram, even though is claimed the opposite. qi, or q_i or Q.i is an abbreviation of input quantity. According to Powers (B:CP, Glossary) a physical quantity (or “phenomenon”) is “A perception identified as part of a physical model of external reality.” And further input function is “the portion of a system that receives signals or stimuli outside the system and generates a perceptual signal…” Even though he himself often talked about “controlled quantity”, I think it would be wise to try to stick to these definitions and situate always qi like all the other (physical) quantities outside the system.

May I please repeat my question: What do you think about my different modelling of the control situation in Steven’s Power law experiment? In that experiment the subject select numbers which correspond to the heard sound strength. In Rick’s model the subject is controlling the difference between sound perception and number perception. In mine (and actually in Mackay’s 1963 model) the subject is controlling the number perception with the heard sound strength as a reference. For me it seems that latter is less complicated.

[amatic]

@EetuP

In the experiment the subject first hears a sound and (then?) sees a number which is said to express or correspond the loudness / strength of the sound. Then she hears a newr sound and is asked to show a number which corresponds the strength of that new sound. For me that setting feels somewhat different from tracking tasks. It feels like the subject is not so much controlling the difference between the sound and the number but rather their sameness so that the perception of the sound strength becomes a reference for the perception of the numerical “strength”. Something like this:

Wikipedia has a nice collection of exponents for the Stephens power law: Stevens's power law - Wikipedia
Maybe length and area could be an easy pair to put in a tracking experiment. Though, it feels arbitrary what is defined as “same strength” or “same intensity”. Maybe it would feed differently while doing the test.

The diagram is missing something between p1 and r to indicate p1=r, or maybe r is not necessary. In Rick’s diagram psi1 and psi2 are subtracted (if you look closely, there is a minus sign after the first psi) to form a new perception or input quantity of the “relationship of intensity inputs” and the reference for that relationship is “sameness”.
From Rick’s diagram: e = qi* - (psi1 - psi2). Yours makes an identical equation, if we assume qi*=0, e = p1 - p2

[amatic]

@MartinT

So far as I can see, this thread about diagrams has not made clear (though Eetu tried) the distinction between what happens in the unknown Real Reality between the final, say muscular, effects on the unknown Real environment and the first affected sensors (e.g. rods, cones, and the equivalents in the other senses).

Personally, I don’t think we as people and as researchers of behavior have any access to the Real Reality. As you say, “we shall never know what it is”. All we know is our experience. The diagrams of control systems are conveying relationships between measurable quantities, and have nothing in them referring to the Real Reality.

Quality of Control, however you measure it, can be only as good as are the relationships between what happens in PR and what really happens in Real Reality as a function of whatever it does to the influences of our control actions on it.

I don’t think this is testable. If we cannot ever know (measure) Real Reality, then we cannot verify that a neural signal comes to be correlated with a Real Reality quantity.

When I want to draw a diagram like those cited in this thread, I always try to show (a) that the input to the organism is an output from RR, which the perceptual system converts to a perception, and (b) the perceptual signal defines the content of the relevant portion of PR.

I find this way of thinking about diagrams consistent: diagrams don’t show inputs to the organisms as outputs from Real Reality. Diagrams show inputs to the organism and outputs from the organism as quantities measurable in the environment of the researcher, outside of the organism. Neural signals are ALSO quantities measurable in the environment of the researcher, but are inside the skin of the organism. There is no metaphysical difference between neural signals and environment quantities, it is just that environmental quantities are more accessible for measurement, and neural quantities are hypothetical. They are proposals for what is happening somewhere in the nervous system if our diagram shows a model in good accordance with behavior, but in principle they are measurable.

[rsmarken]

@MartinT

@rsmarken

When you observe a poor QoC, how do you know that the observed QoC is a result of a low correlation between P and the corresponding variable in Real Reality rather than poor (low gain) control of a P that correlates perfectly with the corresponding variable in Real Reality?

You don’t.

A better question would have been "How do you measure QoC since, according to your view of things, the variable being controlled is in the “unknown Real environment”?

[MartinT]

@rsmarken

A better question would have been "How do you measure QoC since, according to your view of things, the variable being controlled is in the “unknown Real environment”?

The variable being controlled is the perception, and QoC is measured by what in the perceptual reality of an experimenter the experimenter chooses to measure. If the subject is controlling, say, the brightness of a light, the experimenter will discover poor QoC if she measures the loudness of a sound. The TCV is likely to be the best way of approximating whatever perception the subject is controlling “in the wild”. If the subject is supposed to be controlling something like the location of a cursor in pursuit tracking, the computer’s idea of where the cursor and target are is a reasonable way of getting a measure.

All you can know about the Real Reality that leads you to perceive a movement of the cursor is what your perception does when you move your mouse or joystick. That you can know only the the extent that your perception changes in response to your movements. Good QoC, and you have reason to believe something like the function you perceive must be performed in Real Reality.

@amatic

Quality of Control, however you measure it, can be only as good as are the relationships between what happens in PR and what really happens in Real Reality as a function of whatever it does to the influences of our control actions on it.

I don’t think this is testable. If we cannot ever know (measure) Real Reality, then we cannot verify that a neural signal comes to be correlated with a Real Reality quantity.

We can never know how Real Reality produces the effects that we sense and that get translated into our perceptions, but we can say that somehow it does perform something like the functions we identify in what we perceive, if our control is good (or if we perceive uncontrolled things to happen without our control, such as Newton’s apple). Newton guessed that in Real Reality something or other did things that acted to produce effects very like a gravity force proportional to the product of two masses divided by the square of their distance. Einstein observed that these same effects were very slightly more like what would happen if Real Reality actually provided us perceptions of space-time curvatures that we could usually perceive as forces. What Real Reality actually does to achieve those effects is what we can never know.