[Martin Taylor 2014.12.12.23.33]
[From Rick Marken (2014.12.12.1300)]
RM: The problem with this discussion of control of behavior is that aome people have been using PCT in a most unusual way: to explain away rather than explain the phenomenon (fact) of control of behavior.
I take it that this is aimed at my insistence on the idea that control is of the variable that is directly compared with a reference value, namely the perception in PCT or the "output" in classical control theory.
But I suppose it depends on what you mean by "fact" "theory" and "control". With appropriate choice of meanings, I suppose you can talk about "the fact of control" of some environmental variable. For example, you could define an environmental variable to be "controlled" if and only if it varies less under the influence of a disturbance than it "should" (presumably according to some theory about the variable's responses to the disturbance). If you did that, then of course you can control any kind of variable observable by someone who cares to measure the variable and the disturbance, and has a viable theory about how the disturbance "should" affect the variable.
I prefer, and this is just a personal preference, to treat "control" in a more technical sense. Somewhere there is a reference value and a variable that is kept near the reference value despite other influences on that variable, which is said to be controlled. I'm going to call that variable the CV.
Maybe the CV and its reference value are observable, maybe they aren't, but that's irrelevant as to what is actually controlled. If they are, the observer has a good chance to claim observation of "the fact of control".
More often, as in what Rick is talking about as being "explained away", an observer can see something in the environment that appears to be being influenced by some effect, but that is not being affected as much as the apparent influence would seem likely to do. The observer sees nothing that might be a reference value for the thing tha doesn't change "enough", but anyway, the observer has this theory about how much it "should" change if the theory is correct, and if it doesn't then the "fact of control" is being observed.
Now in the _theory_ of perceptual control, the CV is not observable, being inside someone's head, and using the _theory_ of perceptual control, Powers devised a "Test for the Controlled variable" (TCV) to see if the CV or something like it could be figured out using the observation of "the fact of control".
The TCV is simple, but subtle. First, the tester must make a guess about what is being controlled, and what in the environment corresponds to it (this is the putative Complex Environmental Variable, or CEV. Next, the tester must determine whether the probable controller can sense what is necessary if control of the candidate CV is responsible for the effects observed. Next, the tester must determine whether the probable controller could influence the environmental variable that is being observed. Finally, the tester applies an influence (a disturbance) to the candidate CEV, and to various other possible candidate CEVs, and sees which one is least influenced as compared to the amount it would be influenced if it were not being controlled. That one is provisionally taken to correspond to the CV being sought.
The TCV never assumes that the CEV is being controlled. Sometimes, as in some of Rick's demos, there is a clear choice among discrete possibilities (which of these three cars, for example), but often there is not, and Rick has a demo of that too, where the question is whether control is of X+Y, X*Y, or some other function (undefined) of X and Y, when a subject controls size. The outside observer can see the CEV, but unless the outside observer can also see the reference value and the comparison mechanism, the CV is hidden and can be approximated or selected only by use of the TCV.
As Bill Powers as fond of saying, the only fact we have without theory are our direct perceptions. All other facts depend on theory, and that includes "the fact of control", even if it is the subtitle of LCS III.
Martin
