Hi Danko et al
SpreadsheetHierarchy.zip (19.1 KB)
I’ll try to answer your questions here.
Rick: As you note, in PCT higher level control systems set the references for lower level control systems as the means of controlling the higher level perceptions. So the soccer goalie presumably has a higher level system that is controlling for preventing the ball from getting in the net;
Danko: Yes, but goalie may want to catch the ball or kick it ahead. The walker int he park may dodge the ball as you say, but may also choose to catch it or kick.Â
Rick: I agree. These are choices because you can’t do both; the goalie has to either kick the ball or catch it; you have to dodge the ball or catch it. If a person is in such a situation, they are in conflict – they have two different goals for the state of the same variable. A person who is in control doesn’t make such choices; if they did they wouldn’t be in control. A skillful goalie does whatever is appropriate – catching or kicking – to achieve whatever the higher level goal is at the moment. An unskilled goalie has to make a choice (because they are in conflict) and that reduces their control and so they are likely to delay control of whatever variable they end up deciding to control.
Rick: Â It’s true that you have to have options in order to control well. But when you are in control you automatically select the options that achieves the goal; no choosing (in the sense of “deciding” between two or more options) is involved.Â
Danko: I understand most systems are automated and have references set from aboveeâ. But above is the system that picks whether to use arms or legs to achieve the goal, and above is a system that picks the goals for that system.
Rick: What I mean by “automated” is that the outputs that bring a system’s perception to the reference are a deterministic function of error. Those outputs are references for lower level systems when the system involved is a higher level system. So when control is working correctly there is no “choosing” what to do. This is true even when the perception being controlled is a program perception with “choice points”. When you know how to control a program perception, such as a recipe, skillfully, you make the “choices” at the choice points automatically.Â
Â Danko: Maybe there is no choosingâ involved on all levels, but some levels must haave options and pick which system to use. Am I wrong?
Rick: I believe so. From a PCT perspective the only time you are having to choose is when you are in conflict.Â
Â Rick: The quantitative analysis of a control loop shows that the outputs of a control loop are completely determined by the goal (reference) of the control system (the reference signal) and the net effect of disturbances to the controlled variable: that is:
output = reference Â - 1/k *disturbance
Danko I didn’t know about this formula. I know only about qo = Ko * (r-p).
Rick: You get my formula when you solve the simultaneous equations for the control loop for qo. But your equation makes the point just as well. Output (q.o) is determined by error (r-p). There is no choice about what you do (qo) when there is error (r-p). When qo is the output of a control system above the lowest level of control systems, then qo sets the reference (goals) of lower level control systems automatically.Â
Rick: My formula just states that the result of this error driven output process in a closed loop is to produce outputs that vary with variations in the reference signal while compensating for disturbances to the controlled variable.Â
Rick: I don’t know if this will help but I developed a spreadsheet model of a three level hierarchy of control systems that shows how control systems at higher levels automatically set the references for lower level systems in a way that keeps the perceptions of all systems under control. I’ve attached it in a zip file. There is a paper that describes it in my book “Mind Readings” . The demo is a bit abstract but it does show how a PCT hierarchy works. If you can get it to work I’ll be happy to answer any questions you might have about it.
Â Danko: What is your formula saying? Â
I notice if reference is 0 (catch the ball), and there is small disturbance (wind blowing slowly the ball to the right)=0,5, with a constant K=1,Â
Danko: output = 0 1/0,5*1 = 0-1/0,5 =-2 What does this mean O=-2?
Rick: I think you got the math a little wrong. If the disturbance is .5 then the output (with r=0 and k = 1) is -.5. So the formula I presented is just a steady state expression describing the fact that, when control is perfect, output precisely cancels disturbances.Â
Rick: Since the output of a control system in a hierarchy of control systems is the goal (reference) for lower level control systems, it is clear that the goals in a hierarchy of control systems are varied deterministically (per the formula above) – goals are not chosen, in the sense that you can’t decide that the value of the output (the lower level goal) should be x rather than y, for example.
Danko: Yes. Still can a higher level system choose between two systems to use in order to control? Like closing a door. Have you ever kicked the door shot? I have. It seemed like a choice to me instead of using hands.
RM: If you experienced this as a “choice” then there must have been another variable you were controlling that Â made you reluctant to kick the door; perhaps you were also controlling for not damaging the door. I have been in this situation and experienced it as a choice for exactly that reason. The PCT explanation would be this: In most cases, when you push a door to open it with you hand and the door doesn’t open you just push (or pull) harder and harder (because error is increasing). But if you reach the limits on the amount of push (or pull) you can produce and the error is still there (the error being the difference between the state of the door-- closed – and the reference state – open) then higher level systems, possibly controlling for a program perception like “if door doesn’t open then kick it in”, automatically set a reference for a lower level system to kick the door in. But this reference may set up a conflict with another system that is controlling for not damaging property. So now you feel like you have a choice; kick in the door or keep the property undamaged. Conflicts like this are usually resoled pretty quickly through reorganization; how, I don’t know.
Rick: But I will say that controlling not only doesn’t involve choosing actions (or goals); it can’t work by choosing actions (or goals). Control only works when actions (or goals) are a deterministic function of error (as per qo= f(r-p)). Given an error, you can’t arbitrarily choose what to do (qo); if you did, you would not be in control. Â
Â Rick: This gets into the free will-determinism debate. We talk about this a bit in “Controlling People”. It turns out that, from a PCT perspective, we are neither free (whatever that means) or determined. We control. But it’s a somewhat complicated discussion. It might be interesting to discuss this in the mol group. If you think so, feel free to re-post this to that list.Â
Danko: Hope we all get something out oft his without getting into discussion determinism vs. Free will which in my mind is pointless.
Rick I agree!Â
Cheers to you!
And to you, too.Â
Richard S. MarkenÂ
“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers