VS: Loop and comparator (VS: VS: Perceptual Control Theory Talk)

[eetu pikkarainen 2016-11-08]

Dear Bob

Perhaps it is not a total on/off as you wrote earlier? I have in mind the phenomen on of so called “super-normal stimulus”. In fishing the replete pike will bite if the lure is exaggerating enough some features of the normal pray. Also for us human beings
seeing a special delicious food can make us feel hungry. Of course these phenomena are connected to learning / reorganization as you wrote below.

ATT00002.jpg

P.3SimpleLoop_WhatWhyHow1.jpg

image00280.png

ATT00003.jpg

ATT00004.jpg

···

Eetu Pikkarainen


Lähettäjä: Bob Hintz bob.hintz@gmail.com
Lähetetty: 8. marraskuuta 2016 0:44
Vastaanottaja: csgnet@lists.illinois.edu
Aihe: Re: Loop and comparator (VS: VS: Perceptual Control Theory Talk)

bob hintz 11-7-16

I am wondering if the split between Bill Powers and Bill Glasser was over the simplification that removed the possibility of a hierarchy of internal variables that might be parallel to the hierarchy of external variables. Glasser was more interested in
how actors learn to control and/or decide to control some external variable given all the possibilities that might exist in the external environment at any given moment.

When we acknowledge that virtually no perceptions of external variables are continuously controlled (oxygen for humans is the only exception that immediately comes to mind) then starting and stopping a given control loop that passes through the external
environment would seem to be a crucial problem. I stop drinking from my cup when it is empty, but I might choose to refill it. If the coffee pot is also empty, I might choose to make more. If there is no coffee in the kitchen, I might choose to go to the
store and buy more OR I might decide that tea would be fine if I have that available, but why do I want to drink anything at all?

I suspect that only the tiniest loops can be examined without considering reorganization outputs.

bob

On Mon, Nov 7, 2016 at 12:44 PM, Martin Taylor
mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2016.11.07.12.38]

On 2016/11/7 12:57 AM, Eetu Pikkarainen wrote:

Thank you Martin for these clarifications!

I agree that of course the loop must be complete and unbroken to be a loop at all. But I have got a habit to see every interaction - even physical between two objects - virtually circular consisting of one part affecting the other and the other affecting
back (or v.v. – usually at the same time).

Yes, that’s a feedback loop that we call Newtonian action-reaction. There are lots of different kinds of feedback loop. That’s one, a control loop is another.

What is the difference when at least other side is living, is that the living being is not just reacting or affecting back but it is controlling or trying to control its experience, and thus “selecting” (or pitching) its reaction according to its internal
standard or reference level.

Make those last words plural (“standards” and “levels”), and I will agree with you. In my opinion, the existence of perceptual control is what defines life.

The differentiating feature is the comparison of the actual perception and wanted perception – independently from whether the reference value is fixed or changing. But in addition for living beings the reference values are always changing either ontogenetically
or phylogenetically – by learning / reorganization or evolution. That’s why I would like to see at least a virtual “comparator” in every living control system.

Well, there is, even if the “virtual” comparator has no physical existence.

You said that the loop must do a sign inversion from perceptual signal to error signal. I think that this “sign inversion” must be just a mathematical convention, like Boris also said.

No it’s not a mathematical convention. The point is that to be a negative feedback loop, any change at one point in the loop must return around the loop with the opposite sign. If it returned with the same sign, the loop would be a positive feedback
loop, and the value at that point would grow until it arrived at some physical maximum positive or negative value.

There cannot be negative neural currents or signals, can there?

Yes and no, and there you hit on one of a whole bunch of simplifications. Here, rather than address explicitly your comments, suggestions, and questions, I want to talk about simplification, what is simplified and why. “Why” may be easiest to talk about
first.

Even in a very simple organism such as a bacterium, the signalling circuitry, involving chemical and electrical functions, can be too complex to understand in detail unless you are a career specialist in that one organism, if then. So simplification is necessary
in order for anyone to be able to get an idea of the main functions that are happening in respect of some property that interests you, even if all the detailed operations and reactions were known in excruciating detail. Secondly, I don’t believe all the detailed
operations are known for any living organism, let alone a reptile or mammal or fish. So any description of what goes on in their life has to be a simplification, because there is no alternative.

When we talk about perceptual control in a human, the simplification in what we talk about and model is incredible. We discuss a brain that has billions of cell and trillions of connections, and try to create models that do what it does, using models that usually
have fewer than tens or connections. The miracle is that some of these models (most of them being perceptual control models) reproduce observable human behaviour in certain circumstances rather accurately. Some simplifications are explicit and can be discussed,
while some (I assume) we use without realizing that we are simplifying. Here are a few that are relevant to PCT as discussed on CSGnet.

  1. Neural currents (I’ll deal with the “negative” issue separately). We know that the concept of a neural current is a simplification of a precisely timed set of nerve impulses from many different neurons. We simplify the set of neurons that we assert to be
    contributors to the neural current, by representing the whole set or bundle as a single “wire” connector. The neural current flows along the wire in the same way as an electric current flows along a copper wire.

  2. Linearity. When we analyze the response of a control loop to a change in reference or disturbance value, we usually treat the different functions as though they were linear, although we know they are not. Perceptual signals, for example, are much more like
    logarithms of their inputs than like linear summations. Linear equations are much easier to analyze than are nonlinear ones.

  3. Negative neural currents. If a component is treated as linear, then its values must be allowed to go negative. We know that there cannot be fewer than zero impulses from any neuron, so how can there be a negative neural current. We simplify by saying that
    positive current excites whatever it affects, whereas a negative neural current inhibits whatever it affects. So if a comparator (linear) performs the function r-p, it means that the perception inhibits the reference in creating the output error if they are
    both positive.

  4. Noise and uncertainty. We treat the neural current purely as a number. It is what it is. But over time, some changes are caused by inputs to whatever function outputs the neural current, and some are due to the function’s noisy operation, whether it be a
    retinal cone responding to individual photons or a comparator that is really a lot of individual neurons producing pulses that depend on their immediate past history of firing and on the various input pulses. The smoothly changing neural current with a well-defined
    value is a simplification necessary for most purposes, but is important in some other theories of behaviour.

  5. Top-level “virtual comparator”. The HPCT hierarchy may be simplified in its assertion that reference values are derived only from the outputs of higher-level control units. If we accept that possible simplification, then a unit at the top level of the hierarchy
    has no incoming reference signal. Not having a reference value is different from having a reference value of zero, but if there is no actual comparator, the value provided as input to the Output Function of the loop is the perceptual signal. Apart from the
    lack of a sign inversion, this is the value that would be provided if there were a comparator with a reference value of zero. If you want, you could call this effect a “Virtual Comparator”. For any feedback loop to be stable, the feedback must be negative.
    The sign inversion can happen anywhere in the loop. Sometimes it is in the environment (for example pushing on one end of a see-saw rather than the other), sometimes in the output function, sometimes in both. All that matters is that overall there is an odd
    number of inversions around the loop.

6, Ignoring internal feedback loops. As Boris has pointed out, we survive because there are inside us, and inside any living organism, myriads of feedback loops of varying complexity. The perceptual control hierarchy does not simplify them out of existence,
but it simplifies by separating them from the hierarchy into what is called the reorganizing system, which influences the detailed structure of the perceptual control hierarchy. A further simplification is to ignore most of the functioning of the reorganizing
system, other than to say that when “things” are not “working well” reorganization shakes up the connections in the perceptual control hierarchy faster than happens when things are working well.

I think that’s enough to be going on with. We couldn’t easily analyze what happens in even one control loop without those simplifications, let alone the behaviour of a whole hierarchy of control units. But we do find that if we use the simplifications, our
analyses and simulations work rather well. This means that when we try to see what effects the simplifications have, we can use the maximally simplified version as a baseline, not expecting (in most cases) very much improvement from the increased complexity.

I hope this clarifies matters further, rather than making them more complicated for you.

Martin

Does this “sign inversion” mean only that the output effect which it causes, affects negatively or decreases (via mediations) the same perceptual signal which is “inverted” in comparator? You said that there is not a “neural current” in brain because of
its unbelievable complexity. But could we use a most simple first level control loop as a helpful example?

The next one is a pure (non-expert) speculation! The input function is (in) a receptor which senses the intensity of light and the output function is (connected to) a muscle which decreases the aperture of the eye. Probably there must be – in addition to
the reference signal link from above – also an internal reference level (reflex) in this kind of system. When the amount of light and as a consequence the perceptional signal of the light intensity perception increases over some threshold level then the comparator
starts to send a (normal, positive) neural current as an error signal to the output function. Then the sphincter (?) muscle lessens the aperture so that the perception signal would decrease back to the threshold level. Is this example plausible??

(I would like to continue the previous example to a scenario about an animal which lives in a practically perfect darkness but which still for some reason or other has eyes. For that animal the previous fixed internal (and also external from above) reference
level could be zero. That means that as soon as any minuscule perceptional signal of light intensity would exist then the comparator starts to send the error signal to the output function. The higher is the perception signal the higher is the error signal.
So even if there were only one straight “wire” from input function to output function, I would like to call this part of loop comparator, just to tell that it is part of a living being and prone to reorganization.)

OOPS, after writing that above, I noticed that near your really nice diagram you also did say that the sign inversion happens in Output function. But in other places you say that the inversion takes palce between percetional signal and error signal: r –
p = e.

Your comment about the “environment” was really good! It is just because of that complexity of environments that the loose talk about existence of the borderline between organism and environment is problematic. There is a myriad of nested and parallel controlling
loops and they all have their own environment.

At this point I am not yet absolutely convinced about the concept of disturbation, but I must study it more and come back later J

Best

Eetu Pikkarainen

Lähettäjä: Martin Taylor
mmt-csg@mmtaylor.net
Lähetetty: 4. marraskuuta 2016 19:26
Vastaanottaja:
csgnet@lists.illinois.edu
Aihe: Re: VS: Perceptual Control Theory Talk

[Martin Taylor 2016.11.04.12.52]

[from eetu pikkarainen 2016-11-04]

Dear Boris

Thank you very much for this detailed and very helpful reply!

Now I think I see it quite clearly that the focal point of control is the comparator. It is in a way the deepest core of the subject organism.

I would say rather that the focal point of control is the entire loop, no one part of the loop. Control is what in English we call an “emergent property” of the loop. The comparator is indeed a key point in a loop, but only because the reference value (toward
which the controller moves the perceptual value) could be variable. If the reference value is fixed, as is the case at the top level of the hierarchy, no comparator is required but the loop still controls the perceptual value. With a simple “wire” connecting
the perceptual function output to the Output Function input, the loop would still control the perceptual value if the Output Function did the necessary sign inversion to ensure that the feedback is negative.

I think that the problem of “input quantity” (q.i) is connected to the problem of the borderline between the subject and environment. Once we had a professor of psychology who was quite desperately aspiring a scientific breakthrough or even revolution with
his theory that there is no borderline at all. And in deed the borderline is problematic. When you wrote that “We control our perception not limbs.” I got an amusing thought that our limbs are in our environment. The border between me and my environment is
not the skin but it is somewhere near my nerves - and my bones are already outside in the environment.

In PCT, the word “environment” has many meanings. To be precise, you have to say what the environment is of. Is it the environment of the person (observable to other people) or the environment of the single controller. The environment of the single controller
includes all the control hierarchy below its own level. It sends its output directly to its environment, and receives input to its perceptual function directly from its environment – the perceptual signals produced by lower-level perceptual functions. So
your “amusing thought” is correct from the point of view of a controller of, say, the relationship between the rim of a cup and your lips when you want to drink.

Controlling takes place inside the organism

No. Controlling takes place in a loop, and only in a complete loop. We talk about a controller (Perceptual Input Function, Comparator, and Output Function) within the organism, but it does not control anything without its environmental feedback path.

and the functions form the multilayered and permeable borderline between the organism and its environment. Thus we could think that there is something x in our environment (which may be somehow important to the organism like food, or danger) and it causes
a causal effect to a receptor of the organism. Now that effect IN the receptor is the input quantity (q.i). (The quantity of q.i does not depend on the quantity of the x, but rather on the strength of the physical interaction between x and the receptor.) In
the receptor the input function produces the perception signal § which is not same as q.i but a quantitative analog of it.

Correct.

BTW It seems that there is a problem (for me) in that LCSIII diagram which you copied (I have not yet got the book) and which you describe as “But »input quantity« in general sense represent effects of disturbances or effects of behavior or both.” I think
there must be something else too. If I sleep and perceive the comfort room temperature (say +20C) I wouldn’t say that this input quantity consists only of effects of disturbances and/or effects of behavior.

No. It is a reference input, not a perceptual input. You are saying that 20C is comfortable, while lower is too cold and higher is too hot. The perceptual value is the actual temperature, as transformed by your nervous system. It is influenced by things like
your actions (setting a thermostat, opening or closing a window, moving to get into or out of a draft, and so forth) and by disturbances from the environment (passage of a cold front, increasing wind, etc.). There isn’t any “something else” because all the
possible effects are included in “what I do” (my behaviour) and “what happens that I don’t cause” (the disturbance).

If the temperature went higher or lower, then I would call it disturbance and it would perhaps require some behavior form my side.

In PCT terms, it would be not a disturbance, since all the effects not caused by you are caused by something outside, and all combine to form “the disturbance”. You are talking about a change in the disturbance, which “would perhaps require some [* change
in*] behaviour from my side”.

Probably this is only a question of vocabulary, but I would differentiate those effects of environment which do (via mediation of p) cause an error signal from those which do not, and call only the first case disturbance.

Then it is a question of vocabulary. The error signal is always there, whether its value is zero or any other value. Zero error does not mean no behaviour. All it means is that so long as the disturbance doesn’t change and the effect of your behaviour on the
input quantity doesn’t change, your perception will remain as

you want it to be.

If there is a disturbance then the output effect will cancel it by altering the input quantity but that does not mean that this output effect would remove the input quantity altogether, does it?

No it doesn’t. It just brings the input quantity to a value that creates the perceptual value that matches the reference value.

Martin

Best regards

Eetu Pikkarainen

Lähettäjä: Boris Hartman
boris.hartman@masicom.net
Lähetetty: 4. marraskuuta 2016 8:05
Vastaanottaja:
csgnet@lists.illinois.edu
Aihe: RE: Perceptual Control Theory Talk

Dear Eetu.

I answered in your text bellow….

From: Eetu Pikkarainen [mailto:eetu.pikkarainen@oulu.fi]
Sent: Thursday, November 03, 2016 8:27 AM
To: csgnet@lists.illinois.edu
Subject: VS: Perceptual Control Theory Talk

Hi Vyv, thank you for this hint. I see that this last book of Powers is really something to read and I will order it.

Eetu Pikkarainen

Lähettäjä: Huddy, Vyv v.huddy@ucl.ac.uk
Lähetetty: 2. marraskuuta 2016 15:08
Vastaanottaja: csgnet@lists.illinois.edu
Aihe: RE: Perceptual Control Theory Talk

[Vyv Huddy 1247.02.11.2016]

Hi Eetu,

Just sharing an experience of learning about PCT. Carefully reading B:CP has helped but, looking back, I would have learned about PCT more quickly if I’d
started with the book Living Control Systems III. If you are not already aware this book comes with a selection of demonstrations that run on a PC. The parameters of the demonstrations can be adjusted and the effect of these adjustments can be observed and
experienced. This is important because the experience of seeing control happening, and recognising it as such, is key to opening the window on understanding it. For me at least!

All the best,

Vyv

From: Eetu Pikkarainen [mailto:eetu.pikkarainen@oulu.fi]
Sent: 31 October 2016 09:41
To: csgnet@lists.illinois.edu
Subject: VS: Perceptual Control Theory Talk

[from eetu pikkarainen 2016.10.31]

Dear Boris,

Thank you for your relentless criticality, I really appreciate such.

As a newcomer I have problems to follow and find the most essential points – even though I am in a process of re-reading B:CP (in the middle of doing much else). You wrote:

“Rick is definitelly promoting RCT, where »behavior is control«, there is some »controllled variable« in environment and there is some kind of »Controlled perceptual variable« in afferent nerv
going to afferent neuron (comparator). There is no such things in PCT.”

HB : Ricks’ terminology show that »control« is something that is happening in environment. So he is putting all the time »controlled variable« into organisms environment in PCT diagram what
is not the case with Bill. Rick tried many times to make it »official« and I don’t understand how he didn’t succed, because he has Powers lady in his »pocket«. Whatever. PCT is still as it was although Rick is desperately trying to change it with his »Behavior
is control« and »controlled variable« etc.

Bill is not putting »controlled variable« in the environment in the sense that »control« is happening in organisms environment. At least in diagram LCS III. Bills’ defitnion of »control« is
clear (mostly from B:CP) :

Bill also mentioned many times that PCT is about how organisms control.

EP : I understand that 1) behavior (doings) are not controlled and 2) things in the environment are not controlled, but the
third issue is problematic to me. I have understood that it is just the perception i.e. the variable called p i.e. the neural signal created in receptor(s) and transmitted to comparator, which IS controlled. Have got it wrong?

HB : You understand right, that behavior (output) is not control or process of control, so that something or anything in the environment is controlled. Bill is mostlly (99%)
using word »affect«, what I think actually happens in the environment of organism. So yes things in environment if they exist are not controlled but affected. And yes to 3). Only Perceptual signal will be directly »controlled« and that’s why it’s called »controlled
variable«.

So we see that in this Bill’s definition there is no control and if you will read Glossary in B:CP nowhere in other definitions Bill mentioned anything about »behavior is control« except in
term »controlled quantitty« which I answered in conversation between Rick and Martin.

Things in environment are not controlled, because behavior is not control, and there is no »controlled variable« in environment in GENERAL sense that SOMETHING will be controlled by behavior.
Important is that we stick to GENERAL SENSE of PCT so how organisms generaly control. And that is the diagram in LCS III showing clearly.

There is not always (continuously) present something in environment that behavior could control (reduce discrepancy). Generally speaking output (behavior) is used for affecting input and also
for canceling the effects of disturbances.

Bill P (LCS III):

FEED-BACK FUNCTION : The box represents the set of physical laws, properties, arrangements, linkages, by which the acrtion of this system feeds-back to affect its own input, the controlled variable. That’s what feed-back means : it’s an
effect of a system’s output on it’s own input.

HB : Mostly behavior if you will observe yourself don’t have anything special in environment that is »manipulated« by behavior, as behavior is not process of control in general sense. We don’t
dig a cup of coffee with our hands or eat with moving (controlling the movement) of our hands or that we move legs while we are walking and so on. That is self-regulation theory. And Rick is supporting it all the time. He is afterall psychologist like for
example Carver and Scheier (the authors of selfregulation theory based on PCT). When I was mediating between Bill and Carver, Bill told me that they were his students what their book from 1981 prove.

But I think that »controlled quantitty« which could be present in »environment of organism« will sooner or later become part of perceptual signal, which is controlled variable (will be matched
with reference in comparator). And this is the only place where by my oppinion control is happening. Comparator is not a function in the sense all other boxes in diagram are. So it’s the only place where control can happen (nervous system).

Comparator present one neuron or the whole nervous system. I think that all controlling is done in nervous system. I don’t see any place in control loop that
control can be done. Specially not in environment of control system. And perceptual signal is the »controlled variable« which enters comprator. I think it’s obvious.

The result of control is »error signal«.

Here it seems that Bill was equating »Controlled quantitty« with »perceptual signal«. It could be equated with »input quantitty« although from diagram in LCS III is not clear that »controlled
quantitty« exist in any general sense. But »input quantity« in general sense represent effects of disturbances or effects of behavior or both. Bill wanted to create a theory that would generally answer the question how organisms generally control.

In the environment (LCS III diagram) there are present effects of disturbances and effects of behavior (output) added. For disturbances we can assume that they are always present (hit, radiation
of all kinds, sun, etc.), but effects of behavior (ouptut) are present more or less discreetly.

For example in sleeping, what Rick correctly described as »tough one«, existance of »input quantity« in environment, doesn’t have any effects of behavior (output) that would by definition of
control also cancel the effects of disturbances. Disturbance in sleeping for ex. is temperature of the room that is affecting organisms control. In this case »input quantitty« contain just »effects of disturbances« and no counter effetcs of behavior (output)
that could be present in environment. So there is no »controlled quanttity« in outer environment as Rick is trying to present as general principle. But organism is still controlling. The body temperature in accordance with room temperature is controlled (ex.
36,8). So there are »controlled quantities« in organism and »controlled process« in organism that are always present in if you look at definition of control above. So control in organism in general sense is present 24/7 as Bob Hintz denoted this process. I
think he is momentaly one of those who also understand PCT in the sense of my explanation and according to Bills’ definitions.

EP :

The whole concept of “control” is difficult me because of the different theoretical background and also the language: If I try to tell shortly to some of my fellow citizens about my newly found
theoretical inspiration, I have the first trouble how to translate the name of PCT in Finnish! I believe many other outsiders have similar problems. And for you insiders the question may seem so self-evident that it is hard to explain / narrate understandably?
Sometimes I feel like I had understood the whole idea perfectly and then again I am totally confused.

HB :

Beleive me dear Eetu the same thing happened to me. When i was trying to understand PCT I thought that I’ll finish in »mental hospital«. My first »teacher« was Kent (his »papers«
on internet and some meils in 1999). His language is quite clear and understandable, but i thought that the word of author of PCT will reveal to me all »secrets« of PCT. And than troubles began because Bill’s sicentific language in his LCS chain of books is
hard to understand for the begginer. I have to consult with Bill but in the beggining I didn’t understand even his explanations and I was confused as you are now. Then I found Bruce Abbotts’ synopsis. You can get it on this page
http://users.ipfw.edu/abbott/pct/pct.html .
I also returned to Kent and his »papers«. And the »light« above my head start slowly to shine. After all these first difficulties which happened in 6 years I take a break and than one day I started the talk with Rick Marken. Well Rick (somewhere in 2007) in
the beggining understood PCT and as I described many times our converstaion started with which »perception do we control« and not which »controlled variable« we are controlling in environment. So I don’t understand where Rick was lost, but he is totaly out
of PCT line. After that I talked a lot also with Martin and wrote on ECACS. I talked also to Kent and some other PCT’ers. I talked also a lot with Bill whom I finaly understood with no problem.

All in all it took me quite long time to understand PCT. Maybe you could try with the same literature and members I did. But I strongly advise that you don’t try with Rick.
He is a confussion maker because his RCT is contrary to PCT and he has »power« because Powers ladies are fully supporting him (see Barbs’ message). They took me half of my health in last years to keep CSGnet close to PCT. But I don’t know how long will I
persist. Rick is coming with new manipulations and seeking through Bills’ literature to find appropriate answer is not easy job. Anybody can try it.

EP : So how would you define shortly for a newcomer what does the “control” mean? What is it what is controlled and how?
Does controlling take place in a certain place or stage of the closed loop? Or is control rather a function of the whole loop?

HB : I hope dear Eetu, I described you shortly what »control«, »control system«, »ouptut«, »feed-back«, and so on mean through Bills’ definitions. I think this is only fair
approach to hear the »sound« of author. It’s refreshing among all other oppinions. Isn’t it Barb ?

I hope that I manage to present you also what »controlled quantity« and »input quantity« could mean. Problem is that Bill did changed his mind sometimes but Rick is a World
Champion. He is changing his oppinion sometimes daily, weekly, monthly…. Also constant in his oppinion is Kent with his concept of »stability«. It’s PCT. Martin was very stable when Bill was with us. As I understood him he was quite often communicating with
Bill. But now I think that he is too much communicating with Rick and he started to »fluctuate« like him. And he was so stable and perfect in PCT understanding. Rick can cause a lot of troubles in understanding of PCT because he is just relaying on his experiences
which can be misleading as he can observe his behavior as control of limbs. It’s common psychological approach. It’s common sense reasoning. But Bill showed it’s wrong. We control our perception not limbs. Rikcs’ »cancer point« are his demos and tests, because
he thinks that he will understand PCT through them. But as Bill said, the final arbiter is nature. So we should learn there. Experimenting in nature. And I noticed that Rick could be going that way if I may conclude from his latest experiment with Warren.
I hope he will continue his nature experimental work as this is the way to understand PCT.

EP : (And apologies beforehand if I will make follow-up questions.)

HB : As I mentioned mayn times I’m in troubled life position and I don’t have always time. But I’ll answer sooner or later
J

Best,

Boris

Eetu Pikkarainen