In the first attempt at establishing categories in discourse.iapct.org there was a Science category. Like all categories, it had an ‘About’ topic describing its scope and purpose.
The posts now attached to this topic were attached to “About the Science category”. With the reorganization of categories, they became attached to the ‘About’ topic for the Fundamentals category, where they made no sense at all.
Several problems with this “About” notice.
(1) It is not at all clear that “people who would be discussing topics in this category understand that model”. For example, what proportion of the readership would be expected to have a clear understanding of the role of tolerance in a complex hierarchy? What proportion of the readership would clearly understand the role of the external energy supply in reducing the entropy of the CEV?
(2) This description denies the possibility of dealing with the implications of the many different ways two or more control systems might interact.
(3) The presentation and discussion of theoretical analysis is very much an aspect of Science, but it is prohibited in this description.
(4) Discussion of Reorganization is prohibited.
Is that enough to suggest that a total rewrite of the description might be in order? Alternatively, and perhaps easier, the Category title might be changed to “Empirical tests of the HPCT model”, and a new “Science of Perceptual Control” initiated.
Hi Martin
I disagree with all four points you cite as “problems” with the notice describing the “Science” Category. But I do agree that a better name for the Category is “Empirical Tests of PCT”. Then you could have your own Category that could be called “Armchair PCT Science”. I don’t know how to change the name of a category; I think Mak might have to do it. But it doesn’t look like there is any great urgency about making the change since the traffic level at the IAPCT Discourse site is pretty low and CSGNet (my favorite) seems to be making a comeback).
Best
Rick
Rick,
The implication of your final paragraph is that all Science is only empirical, no theory being part of Science. I don’t think many who call themselves “scientists” would agree with you. In my view, all Science is a feedback process like (and indeed “of”) Reorganization, creating new perceptions (in real scientists) along with new ways of influencing the environment. In that view, theory, experiment (also ignored in the description text, though implied by “tests”) and empirical observation go hand-in glove. One is of little use without the other.
Apart from that, could you say in what way you are able to disagree with each of my four points. The text of the category description seems pretty clear and straightforward. In what respect does each of my four points disagree with what the category description says?
Hi Martin
The implication of your final paragraph is that all Science is only empirical, no theory being part of Science.
I certainly didn’t mean to imply that!
I don’t think many who call themselves “scientists” would agree with you.
You means they wouldn’t agree with the implication you derive from what I wrote.
In my view, all Science is a feedback process like (and indeed “of”) Reorganization, creating new perceptions (in real scientists) along with new ways of influencing the environment.
Why not can the phony PCT speak and just say that science is an iterative process of theory development that starts with observation, followed by the development of theory to explain the observation, is followed by observation of the results of tests of the theory, followed by changes to the theory to fit the new and old observations (if necessary), followed by further tests and subsequent refinements of the theory.
In that view, theory, experiment (also ignored in the description text, though implied by “tests”) and empirical observation go hand-in glove. One is of little use without the other.
Yes. The problem is that people often suggest changes or extensions to the theory that are not motivated by observations that cannot be accounted for by the theory in its current state.
Apart from that, could you say in what way you are able to disagree with each of my four points. The text of the category description seems pretty clear and straightforward. In what respect does each of my four points disagree with what the category description says?
Ok. Here’s each of your numbered points with the reason I disagree with them:
(1) It is not at all clear that “people who would be discussing topics in this category understand that model”.
That certainly would be the case. But that would become obvious when they said things about the model that were incorrect, such as:
what proportion of the readership would be expected to have a clear understanding of the role of tolerance in a complex hierarchy? What proportion of the readership would clearly understand the role of the external energy supply in reducing the entropy of the CEV?
People who said things like this about the model without showing how they explain observable phenomena would be politely asked to discuss these things somewhere else.
(2) This description denies the possibility of dealing with the implications of the many different ways two or more control systems might interact.
Not at all. The description of this forum doesn’t rule out the way control systems interact. Indeed, that would be a very welcome topic for research.
(3) The presentation and discussion of theoretical analysis is very much an aspect of Science, but it is prohibited in this description.
Of course not. The assumption that people coming to this forum would know the theory would certainly not prohibit description of the theory. Indeed, discussion of how the theory explains a particular observation would have to be part of a discussion in a science forum.
(4) Discussion of Reorganization is prohibited.
Again, not at all. Indeed, discussion of reorganization would be warmly welcomed.
Why you make this stuff up, I have no idea. But I think it might be because you know that your contributions would not be welcome in this category, if they were anything like your contributions to CSGNet. So go ahead and create your own science category and I’ll see about changing the name of this one to the one you proposed: Empirical Tests of PCT. I’m sure yours will be MUCH more popular than mine.
You are presuming that the category description is supposed to be read as though it was written in normal English, I imagine. If that’s true, how can you justify your comments in this, your most recent contribution?
In respect of your comment on my first point, I must guess your intention in your comment. Are you saying that Bill Powers was wrong in pointing out the critical requirement for tolerance in any control system, or that it is wrong to say that it can have an effect on the behaviour of interacting control systems. Or is your comment directed at the second point? If the latter is the case, could you explain how the energy contributed by the disturbance is eliminated without treating the control system as if it were a simple spring that has no external energy supply?
As for the other three, I can only observe that your reading of simple English is wildly different from mine. That being the case, neither you nor I can say with high probability what implications other readers might take from the words of the description as written.
One last question. As this is supposedly a community forum in which all contributors have equal rights and responsibilities, why do you talk about categories being owned by anyone, particularly by yourself or by me? I personally disavow ownership of any aspect of the forum, other than the content of what I write on it.
MT: In respect of your comment on my first point, I must guess your intention in your comment. Are you saying that Bill Powers was wrong in pointing out the critical requirement for tolerance in any control system, or that it is wrong to say that it can have an effect on the behaviour of interacting control systems.
RM: Tolerance is a phenomenon; it’s not a part of the PCT model. People who understand the PCT model know this; people who don’t, don’t.
MT: Or is your comment directed at the second point? If the latter is the case, could you explain how the energy contributed by the disturbance is eliminated without treating the control system as if it were a simple spring that has no external energy supply?
RM: My comment was just directed at your reference to a CEV (controlled or complex environmental variable). The CEV was never part of the PCT model because it implies that there is always a variable in the environment corresponds to the perceptual variable being controlled. People who understand the PCT model know that that is not the case and they, therefore, refer to the environmental correlate of the controlled perceptual variable as the controlled variable or controlled quantity. This, in recognition of the fact that the correlate of the controlled perceptual variable usually does not correlate with some entity in the environment, an example from B:CP being the “taste of lemonade”
MT: As for the other three, I can only observe that your reading of simple English is wildly different from mine. That being the case, neither you nor I can say with high probability what implications other readers might take from the words of the description as written.
RM: People who understand the PCT model will understand the meaning of my other three responses precisely as I do.
MT: One last question. As this is supposedly a community forum in which all contributors have equal rights and responsibilities, why do you talk about categories being owned by anyone, particularly by yourself or by me? I personally disavow ownership of any aspect of the forum, other than the content of what I write on it.
RM: Since this site is divided into categories with the content in each reflecting the interests of those who would post or read them, I wrote the description of this particular category to reflect the kinds of posts that I would like to see (and not see) in this category. Of course, there is nothing to prevent people from saying whatever they like in any of these categories. But I though one point of this new Dialog Bulletin Board was to divide up discussions into categories so that discussions in each category could be more efficient and to the point.
Given my interests in PCT science, discussions of what the PCT model “really” says or what needs to be added to the PCT model or similar theory first (and only) topics are just not useful to me. I want to see discussions of tests of the model, in particular tests where the behavior of the model is compared to data.
Actually, I see that there have already been several topics discussed under the “Science” category. Some are discussions of model testing, some aren’t. So clearly not everything posted here will be consistent with the description I gave of what should be discussed in this section. So I’ll edit the description of the “Science” category so that says that it is assumed that people understand the PCT model and I’ll just participate in discussions of model testing with people who want to do PCT science. And you can feel free to talk about things like CEVs and tolerance all you like.
Well, it’s certainly interesting to discover what you claim is and is not part of THE PCT Model. All these years, you have not been so explicit, so far as I can remember.
You are quite correct that your understanding of it is different from mine, and since you have decided unilaterally that yours is the only correct model, I guess models based on the various published and CSGnet writings of Powers are ipso facto wrong. As are mine, that depend not only on the guidance provided by Powers, but also on some kind of partial understanding based in engineering of how control must work in the natural world as we perceive it to be.
Tolerance, for example, is very much a part of that model, as anyone who understands either the consequences of using a model that ignores it or who understands what Powers pointed out about it (control loops with zero tolerance are prone to jitter). Tolerance is a region around zero in the “reference-minus-perception” range of variation within which the error signal sent to the output function is zero (or to be precise we should say “approaches zero asymptotically”.
What does your restricted personal model say to the fact that in a neuron-based control loop, variable values below zero are impossible? Does your “correct” model require separated signal paths for cases in which reference-minus-perception has a positive or negative value? If it doesn’t, it doesn’t have much to do with reality, no matter how high a correlation you get between human and simulated tracks in a tracking task. I think we do agree that at this level, the restricted model works very well, but as of today, I’m not sure about that because I think your restricted model is even more restricted than I had thought.
The CEV is, as you yourself have often pointed out, a construction created by a perceptual input function, but you now contradict yourself and say that it isn’t a part of the model at all. The CEV is not a component or property of the real reality through which the environmental feedback path of any control loop through the environment passes. If, however, the properties of the CEV do not mimic pretty closely those of critical parts of the evnvironmental feedback path, control will be pretty bad, so it usually looks as though the CEV is something in the environment that is perceived…
I guess it would be reasonable to generate a topic under a full-scale “Science of PCT” category called “Studies of Rick’s restricted PCT model”. I suppose the heading description of the topic could explain the restrictions, such as that it does not allow for tolerance. Nor does it account for the disposal of the energy in the variability of the disturbance, among just a few of its many problems when used as a map of what might happen either within or outside a living control system.
MT: You are quite correct that your understanding of it [PCT] is different from mine
RM: Always has been; always will be, probably.
MT: Tolerance, for example, is very much a part of that model…Tolerance is a region around zero in the “reference-minus-perception” range of variation within which the error signal sent to the output function is zero (or to be precise we should say “approaches zero asymptotically”.
RM: As I said, it’s a phenomenon (an observed “region”), not a part of the model.
MT: What does your restricted personal model say to the fact that in a neuron-based control loop, variable values below zero are impossible?
RM: That’s an implementation issue. I have dealt with it by building models that control a variable that can be push above or below a reference using control systems that can have only positive signal values.
MT: Does your “correct” model require separated signal paths for cases in which reference-minus-perception has a positive or negative value?
Yes, of course.
MT: The CEV is, as you yourself have often pointed out, a construction created by a perceptual input function, but you now contradict yourself and say that it isn’t a part of the model at all.
RM: No, I’m say that referring to it as an “environmental variable” given the impression that the the controlled variable exists as an entity in the environment. The problem with that is that it leads to the idea that the controlled perceptual variable (p in the model) is a somewhat degraded representation of that environmental variable (which is taken as q.i in the model). This then leads to the ridiculous notion that the TCV can’t tell us what perceptual variable(s) organisms control because the person doing the test only has access to q.i, a degraded correlate of p.
MT: The CEV is not a component or property of the real reality through which the environmental feedback path of any control loop through the environment passes. If, however, the properties of the CEV do not mimic pretty closely those of critical parts of the evnvironmental feedback path, control will be pretty bad, so it usually looks as though the CEV is something in the environment that is perceived…
RM: Yes, as I said, the description of the controlled variable as a CEV leads to the ridiculous idea that what the experimenter sees as the aspect of the environment that is being controlled – the controlled variable – only “looks as though” it is what is perceived. So why not just keep doing psychophysics and forget this TCV stuff? Your concept of a CEV means empirical study of living control systems is a fools errand and, coincidentally, it justifies your armchair (theory only) approach to the study of these systems.
MT: I guess it would be reasonable to generate a topic under a full-scale “Science of PCT” category called “Studies of Rick’s restricted PCT model”.
RM: No, because I don’t want to study the model; I want to study the behavior of those things that the model presumably explains. So under the rubric of “Science of PCT” you can add “Join Rick in the Study of Living Control Systems”. Or, better, just “Empirical Tests of PCT”. I think I"ll add that subhead now and see what happens.
MT: I suppose the heading description of the topic could explain the restrictions, such as that it does not allow for tolerance. Nor does it account for the disposal of the energy in the variability of the disturbance, among just a few of its many problems when used as a map of what might happen either within or outside a living control system.
RM: No such restrictions would be necessary. I just wouldn’t pay attention to that stuff if it wasn’t accompanied by empirical test or demonstration of how a working version of the model that includes those things accounts for some particular behavior or behaviors. So I’ll set up a category called “Empirical Tests of PCT” and see what happens.
I think we have both been looking at this issue of “correct” or “restricted” model wrongly.
If we look at the paradigmatic situation, the PCT model or any other model is a dramatic simplification of what is assumed to be happening inside the organism (or even in the environment). For example, when Powers introduced the concept of a “neural current” he considered it a gross simplification that allowed him to look mathematically at the firings within a “neural bundle” (itself a major simplification) as a “neural current” that he thought might give results within perhaps 10% of observed results when used in a model of control. He reported being pleasantly surprised when the results were better than that.
It doesn’t matter how deep you go into the “neural bundle”, whatever you say about it will be a simplification. For example, in the bundle concept, a nerve fires at some moment in time and otherwise is silent. For the neural bundle concept, the rate of firings per second matters, because the bundle consists of fibres whose firing rates are highly correlated. But at a slightly less simple level of describtion, the polarization of the nerve with respect to the peri-neural medium changes both before and after the so-called moment of firing. Concentrations of Calcium, Sodium, Potassium, and more complex molecules begin to matter; individual dendrites perform what we simplify by calling them computations; individual synapses have varying sensitivities according to the supplies of so-called messenger molecules. And so forth. What happens in Nature is always simplified when we try to talk about it or use some property of it in modelling.
In the same way, what I called “Rick’s restricted model” of PCT and he calls a “correct model” is neither correct nor restricted. It is a model at a level of simplification that Powers (and many others subsequently) found useful in describing to an acceptable precision various effects observed in experiments or empirical observations. I was wrong to think and write about his model as though it was wrong in any respect that matters for what interests him. It’s wrong only when applied to matters that interest some other people, including me.
Rick’s model that he calls “correct” is just a simplification that works very well at one level of describing Nature. As to whether it represents a particular level of simplification on which Powers settled as optimal, that’s something on which I have no comment (at least not in this message).
An important general point about “scientific simplification” is that it is always important to understand that what you are dealing with IS a simplification of something more complicated, often in ways of which you as yet know nothing. Science, over the course of thousands of years, is fundamentally the change of “nothing” to “something”. Science changes the Storm God into a complicated set of models of how energy is transferred in the atmosphere from a source of heat such as the sun or a storage place such as the ocean or fossil fuel into lightning, rain, and wind.
The PCT model that excludes tolerance, for example, is a simplification of one that includes it. Neither is “correct” if your guide in Nature. One may be more incorrect than the other, purely because the less simplified model accounts for more about Nature than does the simpler model, and the added complexity may give insufficiently correct results in the areas it is supposed to address.
In the case of “tolerance” in particular, Bill pointed out that in engineering control, a loop with zero tolerance will “jitter” or “hunt”, the error oscillating between positive and negative. If you aren’t interested in that small discrepancy or it is hidden in other sources of experimental or observational “noise”, it is perfectly fine to use a simplified model without it. Sometimes you can even improve the performance of the model by adding external noise rather than tolerance bounds. But if you want to have fast recovery from a disturbance that changes value step-wise, some tolerance in the loop helps. Likewise, if you are interested in how groups of people can coexist without conflict, you will investigate the role of tolerance.
Rick says he wouldn’t pay attention to “all that stuff” unless there are empirical (in which I think he includes experimental) demonstrations of the effect. Empirical and experimental demonstrations are always important, but they are not always feasible when you get into social situations or into higher-level perceptions. The first explosion of an atom bomb at Alamagordo was described mathematically (and pretty accurately) before the test, and the bomb could not have been built if all the steps had needed to be empirically or experimentally tested. Of course the mathematics had to work with simplified assumptions, just as would any mathematics used to apply PCT to social interactions. It applies equally to supercomputer simulations of, say, galactic evolution and the workings of he brain.
My bottom line here is that there is no primacy in Science between observation (empirical or experimental) and theory. Theory suggests the possibility of tests, but the tests will not be done if they are infeasible or (perhaps) if they might cause irreversible damage. All models are simplifications of Nature, and the only appropriate question is whether a particular simplification answers to the purpose of an investigator.
As one of my professors said about theories: “When two schools of thought each contend that the other is wrong, they are probably both right.”