FW: Feedback Functions (was Giving the Game Away)

[From Bruce Abbott (2015.02.1050 EST)]

Rick Marken (2015.02.15.1315) –

Martin Taylor (2015.02.15.15.55)—

RM replying to me (BA): A backwards way of going about studying living control systems, it seems to me. But you and Martin have never been particularly keen on the approach to studying living control systems laid out by Bill in B:CP and elsewhere. That is, you’ve never seemed to be been particularly interested in developing the TCV as a basis for research.

MT: If you read through B:CP and LCS I, II, III, I think you will find that Bill left a very large number of other ways to study perceptual control and its implications.

BA: The approach to studying living control systems laid out by Bill is not restricted to testing for the controlled variable. It also includes developing generative models whose behavior can be compared to that of living people or animals as observed under specific test conditions. The goal is to understand how control actually is implemented in living systems. The test for the controlled variable will tell you what is being controlled (or a correlate of it); comparing model behavior to observed behavior in real systems can suggest how it is being controlled.

BA: In my collaboration with Bill we did very little testing for the controlled variable; instead, we worked to develop models that might reasonably approximate the architecture of the living control systems involved in such things as feeding and satiety, or the processes of classical and operant conditioning. I’m really surprised to hear you disparaging this sort of work, as if the only thing someone involved in PCT research should be doing is the TCV.

BA: The TCV is of course a valuable technique for identifying variables that an individual is controlling. That said, what can you do with this information once you have it? Many of the variables you control change from moment to moment and from task to task. I want to drive to my office, so as a means of accomplishing that goal I get in my car. Once in the car I control the position of the ignition key until I succeed in getting it into the ignition lock. Then I control the rotation of the key in order to start the car. If the car starts, I control the position of the transmission shift lever, then the position of my foot (onto the accelerator) and on and on and on. All this controlling eventually culminates (if successful) in my getting to the office. Once in my office, I am no longer actively controlling for getting to the office, because I’m there.

BA: Throughout this time I’m controlling thousands of other variables. While driving I’m controlling the position of my body behind the wheel, what I’m looking at, the angle of the steering wheel and of the accelerator pedal. From time to time I’m shifting from controlling the accelerator to controlling the brakes, both as means of controlling the speed of the car. I’m controlling the relationship between the speed of the car and the current speed limit, the car’s position relative to the route I’m trying to follow, and so on. In most cases there is little mystery about what I’m controlling at any given moment and why. But let’s say that we want to establish these things using the TCV. Which activities are we going to submit to the Test? Position of the car in its lane? The route I’m taking to the office? The position of my foot on the accelerator?

BA: Once we get our answer, what can we do with it? Does that information provide any clue as to how these control processes are being implemented in the person’s nervous system? Where there are numerous means to accomplish the same goal, do we learn why a particular means was selected over others? Do we know why this person is controlling a given variable, one that others may not be controlling? Do we learn how those perceptions are being constructed in the nervous system? (In particular, do we learn whether they are constructed in the way HPCT envisions?) Does the TCV reveal the mechanism through which these systems come into being? (Ecoli reorganization? Some more targeted method?)

BA: So what I’m saying is that research conducted to determine what a particular individual in controlling under a specific test condition, while useful for some purposes, only scratches the surface of what needs to be done from a PCT perspective. There’s a lot more to PCT research than doing the TCV.

RM: Really? I certainly have read through them and I never saw anything about other ways of studying perceptual control except of variations on the TCV. It’s all about testing for controlled variables. But I would really be interested in seeing where all these “other ways” are that Bill left.

MT: And if you have read my postings and other material, I think you will find a few references to the analysis of the TCV, which is sometimes useful for a research purpose and sometimes irrelevant. It’s a tool, like any other, most of which you seem to disparage when they are mentioned.

RM: Maybe you could give me one example of an approach to studying living control systems that is useful when the TCV is irrelevant. There is surely one among those that Bill left!

BA: Sometimes the TCV is irrelevant because you already know what variable is being controlled, or because your interest is in exploring the behavior of a particular control system architecture or of changes to the parameters of that system.

Bruce