[From Rick Marken (2011.07.09.0930)]
My “It is the cause” thread seems to have moved to a slightly different topic (the Schouten detection task) so I’ve changed the subject line because I want to get back to what I thought was the original topic: closed loop (control) vs open loop (causal) models of behavior in experiments.
Bill Powers (2011.07.04. 1145 MDT)–
BP: OK, you have a velocity-control system as the lowest level of control. But you can adjust the dynamics of an open-loop model so they exactly mimic the dynamics of a control system model. When the reference
signal is switched from the upper to the lower position, the cursor moves in some particular way – probably a 1 - exp(-kt) curve --to the new position. You can put a filter into the SR system that will do the same thing that your lower two levels of control do, and in the same way. It is always possible to replace a closed-loop model with an open-loop model having suitable dynamics, as long as the independent variables and properties of the environment and output function remain constant. This is why we need the disturbances, because open-loop models
must have a predictable environment free of “unmodeled dynamics” to work. Negative feedback control systems don’t need that kind of coddling.
RM: I just tried filtering the causal model output, using a linear and the exponential filter, and it barely improves things at all f(the R^2 for the causal model goes from .47 to .49, still less than the
R^2 of .62 for the control model). I’ll try some other filtering mechanisms when I get back from celebrating the birth (if not the maturing) of our nation; maybe I’m not filtering it correctly. I’ll give
that ol’ causal model the best deal I can afford;-)
OK, when I use the correct filtering I can get the R^2 for the open-loop model to match that for the closed-loop model. You are absolutely right; it is probably impossible to show the superiority of the closed-loop model of the apparently open-loop behavior in a reaction time task without using a disturbance. I will have to figure out how to do this.
But I think this exercise (in modeling behavior in a reaction time task) was not a complete waste of time. For one thing, it shows that the apparently open-loop behavior in a reaction time task can be modeled as closed-loop control. This is something that I have had a great deal of trouble explaining to reviewers of my papers. What I have done in the reaction time studies is take a task where the behavior seems “obviously” open loop – where the stimulus (the color of the cursor) is not in any way affected by the response (cursor movement) – and shown that the behavior can be modeled as closed loop. This is done by considering that the subject in this task is controlling the perceived relationship between the color of the cursor (the “stimulus”) and the direction of movement of the cursor (the “response”). When the task is “easy” both models account for the behavior in the task (mouse movements ) equally well. But even though there is no basis for selecting one model over the other, the simple fact that it is possible to model the behavior in this task as closed loop should convince conventional psychologists that it is at least worth considering the possibility that the “obviously” open loop behavior they study in their experiments may actually be closed loop.
A second thing about the research that makes it worthwhile (I think) is that it shows that the mechanical use of the causal (general linear) model to analyze the results of experiments can produce poor results because it doesn’t take the dynamics of behavior into account. When the reaction time task was very difficult, the causal model accounts for far less of the predictable variance than the control model because the control model “automatically” takes the dynamics of the behavior into account. The causal model can be made to match the performance of the control model but only after post hoc filtering.
The first point is probably the most important and, I think, makes the effort worth it so far. What do you think? Also, any ideas about how to show that the subject in a reaction time task is indeed controlling the relationship between S and R would be most welcome.
Best
Rick
···
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com