From Bill Powers (2009.09.28.0246 MDT)]
The PCT Persuasion
A persuasion is both a frame of mind and an active process. I mean the title above in both senses. It is time for us in the Control System Group and who communicate by CSGnet to give up internal nit-picking, sniping, and nostalgia for the past, turn our attention outward, and begin a project to persuade all the behavioral sciences that the control system model of living systems is the only correct one among all the others. That claim may not remain true forever, but it is probably the truth as most of those who discuss PCT on the internet see it today. We can let the future take care of itself. Those for whom it is not true are free to observe until they are either persuaded to join in or decide to seek enlightenment elsewhere.
A good many years ago, Tom Bourbon and I wrote a paper called "Models and their Worlds." In it we compared three models as possible explanations of a human subject's performance in a simple task. The first we called the Stimulus-Response model, in which the sight of a moving target caused the response of moving a hand holding a mouse; the hand, mouse, and thus the cursor moved in the same pattern as the movements of the target. The second we called the "cognitive" model, in which the required movements of the hand were calculated according to previous experience with the moving target, and then executed. The third was, of course, the PCT model. We then systematically showed the conditions under which all three theories could correctly predict the observed behavior, then the conditions that would invalidate first the SR model and then both the SR and the cognitive models -- with only the control-system model predicting properly in all three cases.
I propose that those of us who can develop computer models devise more advanced demonstrations along the same lines, expanding this approach (with everyone contributing) by applying it to every current theoretical position held in the behavioral sciences. Those not versed in model-making can then help carry out the persuasion by taking the demonstrations to others of their professional and personal acquaintance and using them to show why only PCT makes sense.
I am proposing, to make this clearer, that we convert CSGnet from a discussion group into a workshop and that we combine our efforts to achieve a goal.
As a starter, here is an idea for showing how both the cognitive model and the SR model can be morphed into the PCT model simply by adjusting a few parameters of a general model. This can actually be done using the "Live Block Diagram" in LCS3, but the morphing process could be made much easier to do. Also, setting up the demonstration as representing a pursuit tracking task will make the distinctions clearer. The object in pursuit tracking is to make the cursor follow a target rather than stabilizing a variable at a fixed reference level.
The live block diagram can be converted to an SR model by setting the reference signal to zero and the feedback function's gain to zero, then adjusting the input gain and output gain and time constant to give a reasonable speed and size of response to the target movements. The SR model will fail when a disturbance is applied to the cursor so its position does not follow the changes of the system's output.
The cognitive model is represented by setting the input function's gain to zero. Then changes in the reference signal will pass through the comparator and the output function, then from the output through the feedback function to the cursor position. If the reference signal is given dynamic patterns of the right kind, the cursor position can be made to vary in correspondence to the target in just the way it does with the control system. The cognitive model will fail when disturbances are applied to the input quantity, because the system does not continuously sense the input quantity. It will also fail when the target moves in a novel way not experienced in the past, since in the cognitive model the required reference signal variations have to be calculated from past experience with the target movements.
The point is not just that the two other models can be made to fail. It's that by adjusting the parameters of the model, either the SR or the cognitive model can be converted smoothly and simply into a PCT model that succeeds in the tracking task for all three conditions. I think this will be the clincher that was missing from "Models and their Worlds." It should show very directly how PCT is related to earlier theories, and exactly what the control-system model can do that none of the existing models can do. Sheer stubbornness aside, I think this will persuade most people that PCT is the next stage in behavioral science.
PROGRAMMERS, START YOUR ENGINES.
Best,
Bill P.
