Sorry Fred…J, I thought that you will answer Rick.
I think Rick that you are a victim of modeling and »thought experimenting« or sitting behind computer and abstracting, rather than experiencing and experimenting in »reality«.
So it’s hard to say that your analysis of Fred’s text is scientific, it looks like more »that you wrote about what you wanted to write« on the basis of model in the brain you created. It seems very similar to Bill’s theory but it is your own construct. So it seems that you created your model without checking it in »reality« or as you said in nature.
I proposed you many times that it’s good to match the model against the final arbiter and verify a »model«, before showing it on CSGnet.
If you would do so, your »model« of driving a car could be more scientific.
As I said. THIRD CLASS »scientific methods« like modeling, experimenting at home behind computer, abstracting, has it’s great weakness. It can give wrong explanations and enable possible rearanging of the »facts«, thus enabling manipulations of the author of the model. Paper can bear anything. But you can’t manipulate »facts« of nature and »Control of perception«. It can cause also terminological mess
So I put my experiences with »real« car drive in the wind, beside your text – on some places. It would bee really nice if in the future you would check your models in final arbiter – nature, so that there will be less possibility that you make misttakes.Â
···
From: Richard Marken (rsmarken@gmail.com via csgnet Mailing List) [mailto:csgnet@lists.illinois.edu]
Sent: Saturday, April 11, 2015 12:42 AM
To: csgnet@lists.illinois.edu
Subject: Re: Protect vs Cancel, etc
[From Rick Marken (2015.04.10.1540)]
Fred Nickols (2015.04.10.0603 EDT)
FN: In the quote I provided Bill wrote about protecting against the effects of a disturbance. In that regard, I agree with you, Rick. We can and do protect against the effects of a disturbance. And, as Bill pointed out in his statement, we do that by canceling those effects.
RM: Right.
HB : Not quite. Can we use term »protect« when any kind of amount of disturbances and output is given ? Or in another words. Do people always »cancel« the effects of disturbances with counter-acting, what is described as »protection« ? Is this the only possibility of the way we make a description of  control ?
FN: On the other hand, I also agree with Boris when he says we can’t protect against disturbances (in the sense of those things out there). We can’t, for example, protect against the wind blowing against our car. What we can do is cancel or counter those effects by turning the steering wheel against the wind).
RM: In PCT the term “disturbance” always refers to an environmental variable that has an effect on a controlled variable. There is nothing to “protect against” if a variable “out there” in the environment, like the wind, has no effect on a controlled variable. So the wind is a “disturbance” only if it affects a controlled variable; if it doesn’t affect a controlled variable then the wind is not a disturbance. So saying that you can’t protect against a variable like the wind that is “out there” makes no sense because there is nothing to protect against – there is no disturbance – unless the wind has an effect on a controlled variable.
HB :
I understood Fred saying : that we can’t do anything if the cross-wind is affecting the course of the car but we can »counter« those effects by turning steering wheel against the wind. Maybe it’s not perfect desciption, but if I understood it, having difficulties with language, I think that you Rick could understood with far less problems then I. But if you didn’t want to understand just to »oppose« Fred’s oppinion, I understand why. We are LCS J. It’s probable that we will understand Fred’s text not »corresponding« to what is written, but to what we wanted in accordance to our references. Â
FN: The driving example offers an interesting contrast. If I’m driving down the highway and the wind is blowing at a steady 20 mph against the driver’s side of my car then canceling or protecting against the effects of that disturbance more than likely results in the steering wheel being turned slightly more to the left than would otherwise be the case. Also important is this: because the wind is steady, the amount of turn to the left remains fairly steady, too, and the amount of drift of my car in the lane is probably close to negligible. Now, let’s suppose the wind isn’t blowing steadily but is instead blowing intermittently in 20 mph gusts. Now, the car starts to drift and I have to correct or compensate for that with intermittent turns of the wheel to the left in order to cancel or correct for the effects of the wind (i.e., protect against drift in the lane).
RM: This is an S-R view of control. The fact is that whether the wind is steady or gusting, the direction of the car (the controlled variable) is always being simultaneously affected by the wind (and other disturbances) and the driver’s output (such as steering wheel direction). Remember p = o + d, all the time. So any deviation of the car from its course is always a simultaneous result of the effects of the disturbance and the driver’s output.
HB : Yes, but simultaneous effects of disturbances and output VARY. And in the case of driver they are not varying simultaneously, so the effects of varying disturbances would be canceled in the same moment. When the disturbances (wind) vary, output is constant (what can you perceive clearly if you focus for a moment on perception of hands on the wheel when you drive the car).
Or as you said :
RM : So while there is, indeed, a delay (called transport lag) between the effect of the gust on car position and the output that compensates for the effect of this gust…
So disturbances and driver’s output VARY with »TIME DELAY, what is causing different effects ON INPUT FUNCTION and CONTROLLED VARIABLE (PERCEPTUAL SIGNAL). If disturbances and output’s counter-action would change effect on input in the same moment that would be really some kind of »protection«. But that is not what is happening in »real« drive.
Bill P :
…the driver must be producing a varrying force that just cancels the sum of the external forces.
HB :
In the moment of »cross-wind« the amount of disturbances vary »d«, while the effect of output stay constant (o). So also perceptual signal (the controlled variable) vary with new quantity of distrubances and consequently »error« signal, after changed perceptual signal is matched to references.
You can see clearly that through the front wind-shield of the car. If you’ll drive you’ll perceive front part of the car being more or less aligned with wanted course of the car on the road with quite steady output. You can check the staeady output with focusing for a momemnt on hands on the wheel). Everything is quite steady. It’s exactly as Bill predicted in his theory. But he described in a little different way as you
Bill P :
…we can observe the driver continually making adjustments of the steering wheel angle, while the car continues in a straight, or very nearly straight line.
HB :
Bill used word »adjustment« for the desciption of control (slight varying perception arround references). And then when the sudden effect of cross-wind (disturbances) »move« or displace actual perception of the course of the car from references), drivers experience more »error«. After experienced »error«, ouput starts with counter action. It is exactly as you wrote. Time delay between different perceptions. We can clearly distinguish in continuity of control »sudden moment« of change of perceptions of the course of the car in repsect to reference, when disturbances (d), wind, vary effects on input.
FN: The word “protect� means to keep safe and it smacks of prevention.
RM: Yes, it means keeping the controlled variable “safely” near the reference specification for the state of that variable.
HB : Why not just keep the controlled varible near the reference. This would be enough. I never saw Bill specifying »safety«.
RM : The control system is continuously acting to keep the controlled variable in its reference (goal) state.
HB : Very good…But II don’t think that people are »continuously« acting. I think that control inside organism is continuous, but actions are more discreet as behavior (output) is just one of the effectors used in continuous control of homeostasis. Behavior (output) is used frequently but not continuously.
RM : ….(and, equivalently, to keep the error signal – which drives these actions – at zero; so the control system is acting to keep itself from acting). In doing this the control system is protecting the controlled variable from disturbance (the net effect of all disturbances, to be precise) – or, if you prefer, it is compensating for the effect of these disturbances to the controlled variable. But this is a side effect of the system acting to keep the controlled variable at the reference (or, equivalently, the error equal to zero).
HB : Perceptual signal (controlled variable) changes with varying in both variables (o+d). You said so. In the case of driver and wind, In the first moment when the effect of disturbances (wind gust) vary, output stay constant. So the action wil compensate the effect of disturbances on »controlled variable«(percpetual signal) with time delay (through comparator). And all this you can easily observe when you are driving in the wheather with wind gusts. Driving is not »protected«, because »error« maybe rarely and more conicidently reach zero. That’s probably why Bill used word adjustment.
Compensating effects follow after the new effect of disturbances took place, because also after comparing perception of »actual position and wanted position« took place.
The »controlled variable« - perceptual signal, was not »protected«, it was »exposed« to effects of disturbances, and muscles activated in accordance with »error« signal, starts compensating the new effect of disturbances on input with time delay. That was caused by variation in disturbance quantity which was compared as »analog perceptual signal« with reference.
So compesating effect came after the new amount (value) of disturbances (gust wind) took place, while the output effects were constant. Check it in »real« drive…
Controlled variable (perception) was not »protected« from varying in values in I.Q. What could afterall mean if we say that »controlled variable« - perceptual signal, was »protected« against disturbances ? It would mean that perceptual signal was protected against effect of disturbances, because counter-action was applayed simultaneosly with effect of disturbances. But how could output knows for new effect of disturbances if it’s happening with no »time delay« ?Â
As you said. There is time delay. And TIME LINE has to be respected as that is what is happening in reality. If that wouldn’t be so, than perceptual signal would not change at all. It would be really »protected«. But it’s varying arround references, as references are. Also references (goals) are not protected from disturbances in organism.Â
This means that nothing is »protected« in the case of driver. It is controlled. And that’s the name of the theory »Perceptual Control Theory«, not »Perceptual Protection Theory«.
FN: You can’t prevent the wind.
RM: Right. It’s just a variable in the environment. When we are analyzing control of the position of a car on a windy day the wind is considered a disturbance because it is a variable that has an effect on the controlled variable (position of the car), an effect that is independent of the effect that the driver has on the same variable.
HB : O.K.
FN: You can’t prevent it from moving your car sideways.
RM: Of course you can. Preventing the wind from moving the car sideways is what you are doing when you control the position of the car. The prevention of sideways movement may not be perfect – especially when the wind is very gusty – but even in a very gusty wind the effect of the wind on the position of the car is far less than it would be if the system’s outputs were not continuously preventing the wind from having as much effect on the position of the car as it would have had if there were no control system acting to prevent it.
HB : How can you prevent effect of wind on the car if you don’t know when it will happen ? And you are contradicting yourself in the statements :
RM: Of course you can. Preventing the wind from moving the car sideways is what you are doing when you control the position of the car.
HB : I understand here that you are saying that you »CAN PREVENT THE WIND« from moving your car sideways, whem you control  This statement is exclusive and unreal. How can you prevent »natural event« from affecting the course of the car ? Are you god ?
RM : …but even in a very gusty wind tthe effect of the wind on the position of the car is far less than it would be if the system’s outputs were not continuously preventing the wind from having as much effect on the position of the car
HB : So in first statement you said that you can absolutely prevent the wind from moving the car sideways with controlling. And in the second statement you said that the moving of the car sideways happens (so you can’t prevent it) but FAR LESS than it would be without control. The final result is that you can’t prevent it. Fred was right. But you can reduce it or limit the effects of sudden change in distrubances (wind gust). But you can’t prevent natural event o having effect on the perceptual control.
By saying that : The prevention of sideways movement may not be perfect - especially when the wind is very gusty … you also admitt that you can’t prevent the »wind disturbances« affecting the position of the car moving sideways. By your words : the more gusty wind the less you can prevent the car side movements.You can’t prevent the actual perception move from reference (»error«). This is what you will clearly perceive if you sit in the car and drive in the final arbiter – nature.
In the AA cars I foumd this quote :
Wind rarely blows steadily, and sudden gusts can catch you out no matter how experienced driver you are…
So I can conclude that you can’t predict when »cross-wind« will have an »effect on the car while you are driving, no matter how expert driver you are. Because it is »sudden« and you can’t predict when it will distrurb your actual perceptual control, no matter how good you are at »continuously« controlling, unless you are a »fortune teller«.
You’ll never predict preciously enough when it will happen to you. That’s why insurance companies exist and they »turn« hudge amount of money.
HB :
You can reduce the effects of the »cross wind« by »continuous controll«, if you :
-
Keep your hands firmly on the wheel.
-
Slow your speed
-
And so on…..
…but you can’t prevent it. There is no »protection« against disturbances in this case, but there is more or less succesfull »compnesation« or »counter-action« which is reducing the effects of »cross-wind«. That’s where experts could have advantage. The problem with succesfull control (compensation) in the case »driver- wind« could be if you compensate with too long time delay, which is caused by other things you could be also controlling.
You can control myriad other »perceptions« in the same time as you are driving, You can talk to your fellow passenger, seek for stations on radio… and sso on. »Multi-control« can focus your attention to other »controled variables« and encreas carelesness and that means that you can end in the ditch beside road, because compensation or counter-action will follow too late in time-line of control events. This is how can you can encrease the unwanted  consequences of effects of »cross-wind«.
Sit in the car and feel all those perceptions and surprises. People are dying because they are careless about wind or if they drive with »split« control. I hope you imagine how it looks like sitting in the car and perceiving the drive. Because that’s what’s happening. All is perception and Bill was right. When you are experiencing nature (real world) only perceptions count.
When I’m talking about »pure« drive it is clearly that driver is controlling for the aligned actual perception of the car in the lane adsjusting it to references, and when disturbances (gust of wind) »move« that actual perception away from the references, counter-action take place And there are time-dealys in perceptions, how ever you take it. You can check it with focusing on perception of your hands on the wheel how counter-action« looks like, not »protection«. When unexpected effect of disturbance took place (gust of wind), you will not counter act immediately but as you said with time delay after matching the actual perception with references, experiencing »error«.
All this events can be experienced in »real« drive. All is perception. When you are in the car, all there will be, are only perceptions. No »control of the position of the car«, no I.Q, and no »simultaneous continuous prevention of the system’s output on the wind from having as much effect….«. Firstt the perception of actual position of the front car in respect to wanted perception of the position of the front part of the car will be changed when effects of wind took place. So the »error« in perceptions will be noticed and compensating-action will follow. The more gusty wind will »bump« into your car, the more »error« in perceptual position of the car you will feel, the more desperate could your actions be. Did you ever see in »real life« what Living Control System do when they drive and something »unexpectedly« happens ? Better you don’t, because you could loose your hair J. But sure their actions are not so elegant as »imagined« in the chair behind computer screen.
HB :
My oppinon Rick is that you are loosing in your abstractions and »abstact« models. So I propose to you that you try every of your models in nature (as you said). In the final arbiter.
Sit in the car, go on the windy street and try it. And I’ll be glad if you try it also with very gusty wind (for example Tornado) how easily you will be »protected«, because you are »continuously« controlling… Â
People as control systems are not so clearly and abstractly controlling when a »real situation« is going on as when they are sitting in the chair behind computer and being smart with their models. Nature is the final arbiter. So go and try and value your models.
FN: You can cancel or compensate for its effects and thereby keep your car in its lane. However, I would wager large sums of money at long odds that your car is not kept in the same position it would have been had the wind not been blowing.
RM: Yes, control is not perfect but it can be very close to perfect.
Bill P.
…the driver mustt be producing a varying force that just cancels the sum of the external forces. Indeed, we can observe the driver continually making adjustments of the steering wheel angle, while the car continues in a straight, or very nearly straight line.
HB : At last we can maybe agree about something.
If you noticed Bill used some different word for »explaning« control of the driver. He used two terms. I think there are many terms that are expressing control in different situation. But I also think that they can be used just in special cases, where one term or another is maybe better. Also »protect« can br be used in some cases of control as suitable. But not in every case.
So overasll term that can descibe what is happening, is not »protection«. So can we make AN AGREEMENT and we use terms as Bill used them moew frequently for describing different cases of control. So terms CANCEL and »ADJUSTMENT« seems to be more GENERALY used than  term »protection«.
By controlling »the position of the car« without any significant variations in amount of »d+o«, Bill used trem »adjusting« the perception of perceived position of the front part of the car with wanted (imagined) course of the front part of the car. In the case of »gusty« wind we can use »cancel«, »counter«, and so on…
It seems to me perfect as each term is describing what is happening with perception. For example : we are adjusting actual perceptions with references. The effet of actions are »adjusting« effect of disturbances in I.Q. and in comparator is adjusted »error« signal.
What do you say ? I’ll break up analysis of your text, and wait for your answer.
Best,
Boris
FN: While I’m at it, I don’t buy the notion of “instantaneous� correction, compensation or protection. I agree that the loop is closed. However, if a disturbance affects the controlled variable at time X then that effect cannot be sensed until time X plus some amount of elapsed time. Even in an electrical circuit where current flows anywhere from half to almost all of the speed of light, that isn’t “instantaneous.� It might be “instantaneous� for many practical purposes but I don’t think it is truly “instantaneous.� Some amount of reaction time is involved in sensing the movement of the controlled variable away from its reference condition and acting to move it back. So, while I get the notion of “protection� and I can agree with under certain stipulations, I’m more inclined to view what’s going on as cancelling, correcting for, or compensating for the effects of disturbances. “Disturbances� are things out there that “disturb� the value of a controlled variable and that change in value is also referred to as a “disturbance.� We need to keep those things straight.
RM: The problem with this analysis is that it fails to take into account the fact that a disturbance has its effect on the controlled variable while the output of the system is also having its effect on the controlled variable. That is, q.i = o + d; the state of the controlled variable, q.i, is at all times a joint result of the effect of the disturbance, d, and the system’s output,o. In your example of the car in a gusty wind, the direction of the car, q.i, is always a joint result of wind velocity, d, and steering wheel position, o. So when there is a sudden, intense gust to the left, the effect of this gust on the controlled variable, q.i, depends on the position of the steering wheel (and the direction it is being turned) at the time of the gust. So the same gust, d, will have quite different effects on q.i depending on the value of o at the time of the gust.
You are mostly dealing here with »external world as the controlled variables are out there. And then you will convert all this »stuff« into perceptual signal. This can be tricky, when you try to turn models into »real« fuctioning. I’ll talk about real controlled variable and that is perceptual signal, because you will hardly find evidence about »pure representation of i.q« in perceptual signal. It could be in some simple cases but not with »complex controlled perceptions«, bur generaly speaking it’s better to dela with real »controlled variables« which clear in PCT. That’s probably where comes the name of PCT.
The difference between i.q. (fake) and perceptual signal (real controlled quantity) can be so huge that is impossible to equatate »i.q.« and p. It can vary very mauch in experiments (psychophyisics) that some generalization, speccialy when controlled perception are very complex is very difficult. But the ratio is not precisley established yet. Maybe some statistical aproximation.
Whether I’m aware of the direction of the car is my imagination. But I’m perceiving the front part of the car and the road and I’m imagining the possible course of the car on the road (reference).
I don’t how you perceive the direction of the car (I suppose that is imagined course of the car on the road), but when I sit in the car I see only perceptions that I can get inside of the car, front shield window, front part of the car, and so on. Sit in the car and you’ll see. Get experiences. Don’t talk only about models. Deal wirh »reality« and real experiences. You said for yourself that nature is final arbiter.
RM: So while there is, indeed, a delay (called transport lag) between the effect of the gust on car position and the output that compensates for the effect of this gust, that output is not a compensating response to the gust (d) but, rather, to the combined effect of the gust plus output at the time of the gust (q.i). That it, the output of a control system is a compensating response to the change in the state of the controlled variable, not to the change in the disturbance… So the causal path is from delta (q.i) (which is delta (d + o)) to delta (o), not from delta (d) to delta (o).
HB : You said it many timeAre you saying again that events in control loop TAKE PLACE IN THE SAME TIME. »…output is not a compensating response to the gust (d) but, rather, to the combined effect of the gust plus output at the time of the gust (q.i). You must be joking. I’m sorry Rick but I call this manipulation.
First change in disturbance and change in input quantity, with constant value of output affecting the sensor (input function) in the same time and then the change in »error« signal. You said it for yourself that there is a lag between this events.
RM :
So while there is, indeed, a delay (called transport lag) between the effect of the gust on car position and the output that compensates for the effect of this gust
What’s else to say more. Time delay between the effect of distrubance (effect of the gust on car) and output that compansate for the effect of this gust is acknowledged. The gust of the wind (change in disturbance quantity) changes the i.q. (when constant output effects are present in i.q.) and that affect input function (time delay) and input function produce perceptual signal (time delay) and perceptual signal is matched to reference (time delay) and »error« signal is produced that enters output function (time delay) and so on and »output compansate for the effect of this gust«. And so on in the loop. Only quantities are changing,
So if the output is causing some specific amount of effects on input (feed-back) through environment and the disturbance specific amount, then the change in i.q, can be caused either by disturbance or by output.
So if there is a sudden change in amount of effects on input function caused by disturbance there is a »time delay« for output to change effects on input in respect to »error« signal which has to change if you want to have some changed effect on i.q. that will cancel the effect of new quantity in disturbance.
So if you’ll sit in the car and drive in wind you will perceive exectly what I experienced : first I perceived the change in the course of the car and then I perceived my hands truning the wheel and after that I percepived »correction« of perception in the course of the »actual perception matching wanted course of the car«. That’s the »time-line«. Andi f you don’t beleive go and try it. The problem wirh your models is that you manipulate with time as it’s abstraction. And »deltas«, what have on the end nothing to do with reality. So stop sitting in your chair behaind computer and start testing your models in reality. Â
RM: In the process of compensating for a change in q.i the outputs of the control system are compensating for the effect that the disturbance is having on the controlled variable. But these outputs are not a response to the effect of the disturbance; they are a response to the change in the controlled variable, a change produced by both the disturbance and the output of the system itself.
HB : You can stop phylosophing Rick. Input function is just symbolic. There is no such a place in external environment that you could »observe« the effects of disturbances and output function. I think that Kent gave a good description »mixed effects«. When one value changes (not to zero) and other stay constant there will be also change in input function and perceptual signal no matter if that is effect of output or disturbance.
If this is not so in delaya, also your statement has no value : a delay (called transport lag) between the effect of the gust on car position and the output that compensates for the effect of this gust… Perfect. This is exactly waht is happening if you sit in the car and perceive driving. First »effect of gust of the wind« and than »
All events in the control loop has to be in accrodance with this two events. No matter what you do and how you model it, it has to »match« the »reality« or nature, whatever.
RM: The analysis of the control of car position that you give above is actually an S-R analysis. It assumes that the effect of the disturbance alone on the controlled variable is the stimulus that causes the “compensating” outputs of the control system. That is, it assumes that delta(d)–>delta(q.i) → delta(o). But delta (q.i) is not solely the result of delta (d); delta (q.i) results from the simultaneous effect of d and o: delta(q.i) = delta(d+o).
RM: When you say that outputs compensate for the effect of disturbances it could be taken to mean that outputs are caused by the effect of disturbances to the controlled variable; in other words, it could be heard as an S-R explanation of how control works. That’s why I prefer to talk about control as involving “protection of a controlled variable from the effect of disturbance” because that’s more like what’s actually going on; outputs are aimed at “compensating” for changes in the state of the controlled variable; they are not aimed at compensating for the effects of disturbances because those effects are mixed with the effects of the systems own outputs.
RM: I think your S-R analysis of the behavior of a control system is a very common way of mistakenly analyzing the behavior of a control system. I think this is true because it looks so much like the output of a control system is a compensating reaction to an observed disturbance, especially when the disturbance is abrupt, as is the case with a gust of wind. It looks like you turn the wheel in response to the gust (after a short delay). This, of course, is the behavioral illusion.
RM: But even though the S-R analysis of control is wrong, I don’t think it hurts anything unless you are a behavioral researcher. It’s a problem if you are a researcher because this kind of analysis makes it seem like studying control systems using S-R type methods is still ok. What’s being ignored when you do this, of course, is the most important research question to be answered when you are studying a living control system: what variables are being controlled by the system (organism)? But for non-researchers the S-R analysis of control, while wrong, probably won’t cause problems,as long as you know that behavior is organized around the control of perceptual variables. Knowing how those variables are kept under control is less important for those doing applied work than knowing that they are there being controlled and that observed behavior is organized around keeping these variables in autonomously variable reference states.
Best
Rick
–
Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble