[Martin Taylor 2008.01.24.12.52]
(jim Dundon 01/24/08.1337est)
I don't think it is usual on this mailing list, though it may be on others with which you may be famailiar, to assume that when someone disagrees with you they are one of: insane, dishonest, or making a joke at your expense.
It might help come to some mutual understanding if you were to assume that none of those are reasonable assumptions. There are other possibilities, one or other of which usually turn out to be correct. Here are a few:
1. Some term is used in different ways by the people who seem to be in disagreement.
2. The assumptions that underly the analysis of the problem differ between the parties.
3. The assumptions are inadequately specified to allow a unique analysis, and the parties have correctly used their analyses to reach different conclusions.
4. One of the parties (or both) has performed a flawed analysis using the same assumptions and logical methods.
None of these possibilities demand that either party is insane, badly intentioned, or mocking.
Now let's consider the question immediately at hand, which I cast as the distinction between prediction and control. I believe that Jim uses both terms in ways that differ from my (and I think Rick's) understanding of them.
[Martin Taylor 2008.01.22.14.52]
(Jim Dundon 01.22.08.1314est)
Jim, with the best will in the world, sometimes I can make neither head nor tail of what you write. This message of yours contains a host of examples.
[From Rick Marken (2008.01.21.1230)]
As I said, prediction and control are not the same thing.
They are inextriably bound. We control parameters in order to ultimately perceive a prediction.
If a weather forecaster says "Tomorrow will be sunny", how is that controlling the weather?
I talked about when we control parameters, not when we don't, And when do, we do it to predict results.
I did understand you to say that you were talking about when we control parameters. My point was to give an example of prediction when there is no connection with control, to show at least that the range of use of the word "prediction" extends outside the domain of control.
Sometimes we make predictions without controling parameters, as in the case you cited. but you really had to carelessly and elegantly overlook the facts to make that comment
Can you be sure of "carelessly"?
24/08
If I control my perception of my posture with a reference value of standing straight, what prediction am I making?
The prediction that you will be standing straight the next second. Good grief man!
Stop pulling my leg with this stuff. I am not gullible enough to believe you really meant those comments.
They certainly are not worthy a clear thinker such as yourself.
I assume you didn't read the expnsions and explanations that followed. I hope I am wrong, and that you did read the rest of my message. Contrary to my usual practice, I'm leaving it in this message in case you didn't.
However, let's go with the flow, and deal with the distinctions I see between "prediction" and "control". Even if you don't use the words the way I do (and I think most people on this list do), perhaps if you can see the way I do use them, you might be able to get a better handle on talking about PCT in a way that would allow others on the list to communicate with you in a way that all parties could understand. It is, after all, a technical mailing list, in which technical terms are used. Perhaps unfortunately, some of the technical terms use words that occur in everyday speech, but use them more precisely than everyday usage. "Perception" is a prime example, but it's not of interest right now.
"Prediction" is literally "saying beforehand". "Control" is acting to make thing happen. There's a distinction between saying something will happen and making it do so. It seems that this is a distinction with which you disagree. As I understand you, to make something happen is to say it will happen.
It seems quite reasonable to say "I predict that if I make the door open, it will be open", but I think most people would think you a bit weird if you went around making that kind of prediction. For example, I might predict that if I go downstairs I will be downstairs. It would be true, but uninformative. One usually "predicts" things about which the person you are talking with might be doubtful.
Now, what about "control". Control starts with wanting some state to exist. Using the preceding example, I might want the door to be open. Perhaps it is already open, in which case, my control is exerted by not acting to change that state. (Parenthetically, quite often on CSGnet soemone will claim that control ceases when the perception matches the reference, but this is wrong. The feedback loop still exists, and the ongoing (non-)action is just what is needed to keep the perception near its reference value. I hope that issue doesn't get in the way of the present discussion).
As I said, control starts with wanting some perceived state to match a reference value, such as that the door be open. But the state in question could be much more complex or abstract, such as wanting to perceive that one's own argument is accepted as being more correct than is the argument of an opponent. Whatever the perception being controlled, if its state does not match its reference value, then a control system will emit output that is likely to have an influence on the perception.
Notice that I did not say "that is predicted to have an influence on the perception" or even "that will have an influence on the perception." I suspect that it is exactly here that we have a communication problem instantiated by
I talked about when we control parameters, not when we don't, And when do, we do it to predict results.
or
... The first thing you learn when you study control theory is
what control _is_. Control is acting to bring a perception to a
pre-specified state and to maintain it in that state, protected from
disturbance. Prediction is just describing what will be perceived in
the future; it does not involve making that predicted perception
happen.
You're crazy. You just said control is "acting to bring about (make happen) a pre-specfied state" followed by control "does not involve making the predicted perception happen" Like in maintainig a standing position?? Only another crazy person would call that blatant contradiction elegant. I know I'm not a trained scientist but give me a break. Stop insulting what little intellegence i do have. Do you ever read the stuff you right?
I'm not going to insult your "little intelligence" if I suggest you re-read the paragraph on which you appear to be commenting, am I? If so, I apologise.
I wil assume you have re-read it, and that you have not changed the way you would respond.
Of course, I cannot assure you that I'm not crazy, because that probably isn't something I would know. What I do know is that when I read the paragraph in question, it makes a very clear distinction between predicting and controlling. To you it doesn't, and it's up to both of us (or to me, if you don't want to participate) to figure out where the problem lies.
So, let's consider your response (I'll omit the "You're crazy" bit, which doesn't really help me understand the problem).
You misquote: The paragraph says "Control is acting to bring a perception to a
pre-specified state and to maintain it in that state, protected from
disturbance." Note that word "Control" at the head of the sentence.
Then, to contrast it against "control", the paragraph goes on: "Prediction is just describing what will be perceived in the future; it does not involve making that predicted perception happen." Note that word "Prediction" at the head of the sentence.
Now consider what you say: "You just said control is "acting to bring about (make happen) a pre-specfied state" followed by control "does not involve making the predicted perception happen". You substituted "control" for "prediction" in the second half of your sentence, thereby creating for yourself the "blatant contradiction" about which you complain.
Back to the main thread.
I just said: "Whatever the perception being controlled, if its state does not match its reference value, then a control system will emit output that is likely to have an influence on the perception."
Here we run into the question of reorganization. There are two question:
(1) Why is it likely that the output will have an influence on the perception, and
(2) does the control system need to "predict" that this will be the case before acting?
The answer to the first question (in conventional PCT) is that if the actions of a control system do not influence the perception, or influence it in the direction of increasing the error, then it will change its actions until they do influence the perception in the direction of reducing the error. That change of actions is, in part, the domain of "reorganization". It's a trial and error thing that on CSGnet is often called the "e-coli" approach to optimization. At very low levels such as muscle tensions, the actions are on the real world, and the laws of nature are, we think, pretty well fixed over the span of many lifetimes, as are things like Newton's gravitational constant. Accordingly, the "e-coli" reorganization can be done over an evolutionary time scale so that each individual does not have to learn by itself how to tense or relax a muscle.
It is clear that the consequences of sending a reference signal to the muscle control system are pretty predictable. Does that mean they are "predicted"? In my use of the term, I would say "no" for the person who is doing the muscle tensing during, say, a walking step, but "yes" for a physiologist telling a class what will happen if he applies a voltage to a frog's leg on a lab table. Am I right to think you would say "yes" in both instances?
One of the aphorisms sometimes used in PCT discussions is "Many means to the same end". Suppose you are controlling a perception of the door's state, with a reference value of "open". You pull on the door handle. Are you predicting that pulling on the door handle will lead to the "open' state for the door? Maybe, but maybe not. The door doesn't move. Now you turn the door handle and pull. Are you predicting that the door will now be open? The door doesn't move. You notice there is a "key" in a "lock". You turn the key and then pull the door handle. Are you predcting that the door will be open? The door doesn't move. You notice that the door frame overlaps the door, and you push the door. Are you predicting that the door will be open? The door doesn't move. You remember that you turned the key, so you turn it back again, and push the door. Are you predicting that the door will be open? You turn the handle and push the door. It opens. Were you then and only then predicting that the action would result in the door being open? Are were you predicting all along, for every action: "If I perform this action, the door will be open"?
My own answer to this is that I would not have been predicting, but that I would have been hoping that each of these actions, all created by the output from the system controlling the perception of the door state, would reduce the error in my perception of the door state. I would have tried many different things, all of which had at some past time been successful in getting the door open. But in the end, it might have happened that none of my actions succeeded in opening the door. What then? If it was important that the door be open, as the only way to bring a higher-level perception to its reference value, I might try things I never tried before, at random or with consideration of actions I might have seen other people take. But I might not. I might try different ways of affecting the higher-level perception (which might be to perceive myself as being out of the room), such as going out of the window.
The point of all this rigmarole is simply to try to ask how you would interpret "prediction" in this kind of scenario. If you say that all of the times, you would be predicting that the action would reslt in eliminating the error in the "door state" perception, then we know that we use the word "prediction" differently, and that this is the source of the difficulty at the point where I joined this thread.
The example might, perhaps, also contribute to the question with which you started the thread, about the HPCT hierarchy leading to robot-like behaviour. But that's not for this message. I dealt with it a little in the earlier message, in the context that you can control another person, provided that what they do when you are controlling them is of their own free will -- which sounds like a contradiction, but isn't, if you read the earlier message.
Does this help us get any closer to understanding one another?
Since you didn't comment on the rest of my earlier message, I leave it appended.
Martin
···
All of your computer simulations are involved in making your predictions happen.
Huh??? Say what???
Computer simulations show the consequences of making certain assumptions and setting certain parameters. What predictions is the modeller "making happen"? If a simulation model with certain assumptions and parameter values happens to produce patterns very like what happens when you ask a subject in an experiment to do something, how does the model make the subject do it?
Your comment sounds very parallel to a hypothetical comment that would say a mathematician solving an equation is "making the result happen". There is a sense in which it would be true, but it's not a sense most people would associate with "making happen." The equations imply the solution, just as the model structure and parameters inply the behaviour of the model. If you are able to predict how the simulation will behave because you can solve its equations, good on you, but just how is that "making your predictions happen?" If the model behaves very much as a subject later behaves, what is making what happen?
In science, we predict results but we don't control them.
Controlling the result of a scientific experiment -- making what is
predicted actually happen -- would be an example of scientific fraud.
You can't be serious!
Why not. He is absolutely right. It would be fraud if it were done consciously.
Especially in psychology, there's a problem called "the experimenter effect" that's been known for a long time, and against which effective experimental design is required: that the experimenter has a prediction of how it will turn out (or worse, a desire to see it turn out a particular way), and unconsciously (not deliberately) does things that make it turn out that way. It's why we have "double-blind" studies. Failing to ensure that the experimenter cannot control the result is a sign of a failed experiment -- one in which nothing is learned, because for all you know, the result might have been the experimenter's doing.
I know its just poor english not poor thinking, but "making what is predicted actually happen"
would be a success. Lying about what happened would be fraudulant.
Oh, I am so glad you aren't a scientist. You would be working for a big drug company and making a lot of money telling Congress and your advertising department how wonderfully your drug performed in the experiments that you controlled so as to give the result your bosses wanted, rather than ones that tried to find out how well the drug works in practice.
OK, Now what? can you extend this to say whether or not a particular
person will come to work tomorrow, kill tomorrow, love tomorrow, be
tomorrow?
In practice, no, because, in order to predict control behavior in
natural situations you have to know many things that are themselves
only poorly predictable.
More to the point, PCT tells you only that if you know some perception a person is controlling, as well as the reference value for that perception, and the gain and time-course of the feedback loop, then by disturbing the controlled perception you can induce ("predict" if you do it as a thought experiment beforehand) certain actions.
I can control a friend in the room with me, under some circumstances. For example, if I am sitting down and they are near the window, I can say "Would you mind opening the window" and more often than not they will do it. I can predict this will happen, and I can control my perception of their action, though not always successfully.
Why is this, according to PCT? I assume that the friend is controlling a perception that I am comfortable, and I assume that they are not controlling a perception that requires the window to be closed (because that would induce a conflict in my friend). I act (speak) so that they perceive me to be uncomfortable in a certain way, causing their controlled perception of me being comfortable to deviate from its reference value. They bring their perception of my comfort level back to its reference by opening the window. If my assumptions are correct, so will be my prediction that they will open the window. And I will have controlled my perception of my friend's actions.
There will be occasions, though, when my friend will not open the window. One of the conditions under which PCT suggests this will happen is if the friend perceives that the perception I am controlling is not to perceive the window to be open, but rather to perceive that I can control my friend. Many people have reference values to perceive themselves as acting under their own "free will". If they perceive themselves to be being controlled by others, this causes an error, and they will be likely to act otherwise than what they perceive the controller to want. So, maybe my friend will not open the window if my request is perceived as being due not to a wish to have the window open, but to a wish to control my friend's actions.
PCT says you can only predict someone's actions if you know all the parameters and all of the controlled perceptions, as well as all the relevant disturbances and fluctuations in the feedback path such as something getting stuck -- which includes whether a person's attention might be distracted at a given moment. PCT says that most social interaction does consist of controlling other people's actions (I'll get disagreement on this claim, I suspect, but it's true in the sense that getting my friend to open the window is controlling her actions).
When you are working for someone, or submit to military discipline, then you are likely to have a reference to allow yourself to have your actions controlled by certain other people, but this isn't the usual case. As I said above, most people like to perceive themselves as acting of their own volition. The friend who opens the window does so of her own volition because she wants to perceive me to be pleased. That in no way changes the fact that I control her actions in opening the window -- it is what enables me to exert that control.
It sounds like you are saying if everyone relates in this way with my
apparatus we can predict performance. So the prediction depends on
everyones cooperation in relating that way.
What it says is that we assume that when someone agrees to act in the experiment, they also set reference values for various controlled perceptions involved with doing what they are asked to do. That's what is meant by "cooperation" in PCT.
That is a relational frame.
This needs explanation. It might make sense.
That entire frame is a refernce signal Bill has established.
Huh??? This makes no sense. I think you are using words from PCT in a way unrelated to their meanings in PCT. But I don't know, since I can't interpret it in context of anything else in this thread.
I don't see what you are trying to say.
But you do. Your following statement describes the relational frame
All scientific prediction
requires that we know the state of many variables that the prediction
model says are involved in the behavior to be predicted. You can
correctly predict the time it will take a ball to roll down a plane
only if you know the state of variables such as the inclination of the
plane, the frictional resistance of the surface of the plane and ball,
weather conditions, etc. If you've ever been in a physics or chemistry
lab you know that you only get the predicted results if all relevant
variables "cooperate" (are carefully contrived) to be in the "right"
states.
How did they get that way?
The experimenter set them up, or else measured them. I don't see what Rick's statemen has to do with anything I would associate with a "relational frame". Clearly, if the inclined plane is at a very shallow angle, the ball will roll down it more slowly than if it is almost vertical. If in a chemistry experiment you add ammonia instead of hydrochloric acid, you won't get the predicted results. All Rick is saying is that you have to know that kind of thing before you can predict how fast the ball will roll, or what will happen in the chemistry experiment.
Likewise in PCT prediction, if you don't have a pretty good guess as to what the subject might be controlling, you won't predict very well how events may disturb his controlled perceptions, and still less will you be able to predict the person's action to correct errors in the controlled perceptions. Because a person agreed to be in the experiment, the experimenter assumes a lot about what perceptions the subject is controlling, and therefore can make quite a few predictions about the effects of disturbances the experimenter will deliberately introduce to perceptions assumed to be controlled.
Bill has often ridiculed the use of statistics in quantum theory. he "didn.t
like it!" Here he uses statistics to support his theory.
I think this is a non-sequiter. In PCT we question the need for
statistics in psychology because we have found that behavior is a lot
less variable -- and a lot more lawful --
What laws are you talking about?
Laws of Nature. Consistency.
If this is true, then you are admitting to the model influencing the behavior and that would belie what Bill has required of his theory.
Nonsense. At least, it's nonsense if what you mean is what the plain English of this seems to mean.
Namely that it be true at all times and in all places. That would include times and places where control theory is not known,
Yep. And it includes life forms incapable of "knowing" anything about PCT, such as flowers, insects, and bacteria.
and be independant of the level of lawfulness.
What on earth does this mean?
That would also be scientifically stated.
What would?
PCT conversations always seem to be contradictory in this way. I am still waiting to hear whether PCT is about all behavior as some of you say some of the time or all behavior as some of you say some of the time..
Is it about X or X?
My answer is that it is about all behaviour of all living things, all the time. Is that what you meant to ask? That is actually its domain of reference.
All behavior would include insanity, murder, infanticide, genocide, suicide, shizophrenia, catatonic schiziphrenia.
Yep. Also tree growth, rutting behaviour of male moose, bacterial movement, ant suicides in protecting the nest ...
In my relatively uneducated opinion, It is not very scientific after decades of dialogue to be unsure and unclear about what you mean .
So what DO you mean? I know language can never make it precise, but your language seems singularly impenetrable. It may be sure, but it's definitely not clear.
Spit it out. "All" behavior or behavior "modified by incorporation of PCT."?
What does "incorporation of PCT" mean? And how would it "modify" behaviour?
I do find that my understanding of the quirks of other people's behaviour (and my own) seems to be clearer from having some understanding of PCT, and perhaps I treat people more tolerantly for that reason. But somehow I don't think this is the kind of "modification" you mean.
Make up your minds whether you want to sell as one or the other or both!!
One or other WHAT???
And be consistant
I think what you want is for PCT to be omnipresent.
Oh, it is, it is! That's unavoidable. But only in living things. Luckily.
I think. Words are what we use to try to
communicate ideas. But I think the best way to understand a theory is
to see how the quantitative relationships actually work. I see this
best in computer programs; people who are better at math can see it in
equations. But words will always be necessary to describe what the
programs and/or math show.
I would suggest that the math is prescribed by the theory and will limit, bind, within the frame of the terminology and methods, the results. In this case PCT. I see it as a creation/discovery maintained by your commitment to your creation.
Finally we come to a paragraph that I think I understand. It's a much argued point in philosophy, with which I happen to agree. It really is a weak version of the Whorfian hypothesis. The strong version is unsustainable, but the weak version may well be true -- what we perceive is much influenced by how we talk about it (maths is talk), as well as vice-versa.
Your post makes it clear that you are strongly opposed to PCT.
You are mistaken.
I am opposed to your assumption that it is the only reality.
Has anyone suggested PCT is the "only reality"? I don't think anyone has suggested that PCT could have predicted the surplus neutrinos that were observed over a particular few seconds in 1987 because of the decay of nickel(?) created in the supernova explosion. They were a reality, too, and their timing proved that if neutrinos had any mass, it is incredibly small (of course, now we understand that neutrinos do have mass, and that mass is incredibly small).
I will continue to attempt to make myself clear by being more direct.
That would be most welcome.
I have attempted to point out weaknesses in PCT
Again, that would be most welcome, but it MUST be based on a deep understanding of the hierarchy of PCT flavours, from basic PCT to generally accepted PCT, to speculative PCT. If you can find weakness in basic PCT, you will have done a service to all of the physical sciences. If you find a weakness in speculative PCT, you may help to strengthen it. So please, please, do analyze and point out weaknesses in all flavours of PCT. In doing so, though, it would be nice if you referred to a version of PCT that its practitioners would recognize.
Please do it right. There's nothing a science needs more than cogent criticism.
I am not opposed to PCT . I am not even opposed to your presenting it as a behavior modifier.
Who has ever done that? And if someone did, why should it matter? Psychiatrists have long looked for effective behaviour modifiers. I'm not sure whether MOL counts as a behaviour modifier, but it does represent PCT in a psychiatric environment, and if it works to make people feel better and to interact better with the world, surely that's a good thing, isn't it?
I AM opposed to your denying that you are presenting a behavior modifier and saying that it is pure science when in fact it is an exhortation to responsible behavior, something done by many others not in the guise of science.
As I mentioned above in connection with myself, it is indeed possible that an understanding of the science of PCT may induce someone to act more responsibly than they otherwise would. But that's not a problem, any more than it is a problem that a knowledge of thermodynamics helps engineers build more efficient engines than they otherwise might. It doesn't make thermodynamics or PCT any the less pure science.
What is your issue, here? Is part of your definition of science that it should not apply in the everyday world?
Martin