Research and Modelling

[From Rick Marken (960210.1115)]


Available data does provide guidance for further _research_; but it
provides little if any guidance for formulating control models

Bruce Abbott (960209.1630 EST) --

When you begin testing for controlled variables, it helps to have some
particular set of observations in mind that one is attempting to explain by
means of a model.

As I said above, when you _begin_ testing for controlled variables it
_does_ help to have some set of observations in mind. But the value of
such observations is _not_ that you are attempting to explain them by
means of a model; the value of such observations is that they hint at
possible controlled variables; they provide you with the _initial
hypotheses_ to test when you do the test for controlled variables.

There is nothing wrong with showing that a particular set of observations
can be explained by a control model. It's just that such modelling can be
quite misleading unless it is clearly the start of research aimed at
determining whether a variable is actually controlled and, if so, what it is.
For example, I could write a control model of the patellar refex right now.
The model could work by controlling a perception of the angle between
femur and shin because this is what _appears_ to be controlled. The model
would be designed so that it kicks up briefly when a disturbance to this
angle is applied and quickly removed. Building this model would be a
nice exercise but I still have to see whether I've got the controlled
variable right (which, of course, I don't); in other words, I still have
to do the reseach that I would have had to do to construct the proper
model in the first place.

I certainly want to move to the lab work as quickly as possible , but have
been facing certain constraints that have made this difficult to carry out.

Then why not do equivalent operant experiments with people? Then you
will be able to test to see what is actually being controlled and develop
control models of this behavior right in the privacy of our own PC.

I think the advantage that rats used to have over people as subjects -- the
fact that rats were general more greedy, selfish and mean than people --
has largely disappeared in the last decade in the US;-)

I don't see this [modelling] effort as a substitute for the lab work, but as
preparatory to it and as something that can be done until the lab gets back
on line. It sure beats just sitting on my hands.

I certainly agree with that. I think your modelling efforts would be
particularly superior to sitting on your hands if you would describe the
kinds of research that should be done to test your models.