Irrelevant side effects

CSGnet Archive CSG1995
1

[From Rick Marken (950322.0800)]

Bruce Abbott (950321.1900 EST) --

although I agree with you that the phenomenon of stimulus control
is essentially a side-effect of control, I don't see it as "irrelevant." If
you wish to understand what someone will do when conditions
change, you will have to learn what effect this disturbance has on
controlled perceptions, and what means the person has at his or her
disposal to correct that disturbance.

And how the means the person has available to correct that disturbance
actually affect the controlled variable at the instant that the person is
affecting that variable and ...

In other words, we have to know many things we usually don't (and can't) know
if we want to understand what a person will do when conditions (disturbances,
feedback funcitons) change. That's why we usually don't care what a person
will do in response to a particular disturbance. All we want to know is what
variable(s) the person is controlling. Once we determine a controlled
variable (usually by looking for lack of effect of a distrubance on the
variable) we can assume that the person will do what is necessary to keep
that variable in a reference state.

The "stimulus control" observed in response to certain disturbances of a
controlled variable is as irrelevant to control as the name written by the
person controlling the knot in the rubber band demo. The fact that "stimulus
control" is occuring is irrelevant to the controller and it is irrelevant to
the person observing the controller because it tells the observer nothing at
all about what is being controlled; or, for that matter, whether or not
control is even occuring.

Did you say that your demo is a hypercard stack? I hate to admit it,
but I DO own a Mac. How about posting the stack?

Hey. I hate to admit that I own a PC;-)

I'll tidy up the stack and post it in a few days. I like this demo alot, if
I do say so myself.

I doubt that many in EAB would find your demo surprising.

I think it is absolutely impossible to surprise or astound conventional
psychologists with any of our demos. I thought my "mind reading" demo would
blow the lid off conventional psychology. But nooo. I presented it to an
audience of cognitive psychologists at my old graduate alma mater, for
example; they thought it was cute but no problem at all for conventional
psychology.

Yeah. Right.

I imagine that a clever enough reinforcement theorist would be able
to "account" for the effects of your proposed disturbances, although it
is easy to see (to me anyway) that the control model elegantly handles

these effects without additional assumptions.

The "clever" reinforcement theorist will simply be BSing; he has no way to
build a working model of the behavior observed in my demo because that
behavior is a side-effect of control and reinforcement theory is NOT a theory
of control. (That, by the way, is why I keep wincing when you talk about PCT
as an _alternative_ to reinforcement theory. How can PCT be an alternative to
a theory that doesn't explain control? It's like saying that Newton's laws
are an alternative to astrology; control theory is a working model of a real
phenomenon; reinforcement theory is a lot of blather aimed at explaining the
illusion of environmental control).

I'm not sure that your modelling approach to explaining "stimulus control"
will be any more convincing to our clever reinforcement theorist than the
results of my demo -- though I certaintly hope it is. Of course, once you
complete the model it will behave just like a subject in my demo; the model
will show "stimulus control" given the first type of disturbance and no
stimulus control given the second. That is, "stimulus control" will be an
irrelevant side effect of the controlling done by the model. But the "clever"
reinforcement theorist can just say "yeah, but that model is a reinforcement
model -- it gets reinforced by seeing a particular pattern of cursor/target
relationships; these relationships cause the model to respond according to
stimulus control or not". Yeah. Right.

I guess my bias is to start explaining control theory to reinforcement
theorists in terms of _phenomena_. The control model is not what causes the
problem for reinforcent theorists - - the control model has been around for
years, it's pretty simple, it's well known to reinforcement theorists and
it's almost always applied to behavior incorrectly. What causes the problem
for reinforcement theoristys is the observation that consistently produced
behavioral results - - bar presses, figure eights, spoken words, etc -- are
controlled results: in other words, the problem for reinforcement theorists
is that behavior IS control, not in theory but in FACT. Reinforcement
theorists don't even know that the phenomenon of control EXISTS! My demo
shows that the phenomenon of control DOES exit -- the subject is clearly
maintaining a logical variable in a reference state despite varying
disturbances. It also shows that one possible side effect of the phenomenon
of control is the appearance of "stimulus control".

I hate to say it, but you're starting to sound like B.F. He thought his
observations were "theory-free," too. (;->

You're right; my observations of behavior are not theory free. They are
filtered through well established physical models of reality. My
demonstration of "stimulus control" is theory free only in the sense
that it is free of any psychological theory. You don't need control theory
in order to see that the subject in my demo is controlling the relationship
between cursor color and cursor position; that relationship would NOT
continue to happen (based on what we know of the physics of the situation) if
the subject's outputs were not systematically being varied in order to keep
it happening. HOW that controlling happens is another question (which you
are answering with your model); but there is no question THAT it is
happening; the subject in the demo is controlling and, depending on the
nature of the disturbance to cursor position, there can be the appearance of
"stimulus control" or not.

Good luck at the meeting, by the way. I'm hope that you will be able to
convince all those EAB types that they have been studying irrelevant side
effects of controlling for the last 50+ years. But they're all probably
looking forward to learning that. Yeah. Right;-)

Best

Rick

2

[From Rick Marken (950328.0920)]

Bruce Abbott (950327.1750 EST)

Perhaps a more complex model [of compensatory tracking] is not needed at
this stage

Since the existing, simple model typically accounts for 99% of the response
variance, I'd have to agree; a more complex model is almost certainly not
needed at this stage.

Bruce Abbott (950327.1950 EST) --

Rick, this [space invading] research has been around so long that even high-
school students are doing it for their science-fair projects.

I know about it and it's a good start; too bad it's only done by high school
students. I think this research never went anywhere because the experimenters
never realized that there are many variables, besides interpersonal distance,
that people control. Indeed, they never realized that the variable
controlled in the "distance control" studies might not be "distance" nor did
they realize that it was a perceptual variable whose reference state was
adjustable; the reference for distance can obviously be changed in order to
control other variables -- if you catch my drift;-)

I'm not yet clear on what the difference is between a "side effect" of
controlling and an "irrelevant side effect" of controlling.

The actions of a control system produce many effects in the world; for
example, the heater in a temperature control system makes noise, affects
humidity, etc. All these effects are irrelevant (from the point of view of
a control system); the only relevant effects of actions are those that affect
the controlled perceptual variable -- temperature at the sensor. So all
uncontrolled effects of the actions of a control systems are irrelevant side
effects.

Sometimes I leave out the qualifier "irrelevant" because, in a hierarchy of
control systems, the irrelevant side effects of one control system may be
relevant to (have an effect on the perceptual variable of) another control
system. An irrelevant side effect of eating soup is the slurping sounds that
result; those sounds, however, might have an effect on the perception
controlled by another system -- the perception of "being a slob", which is
being kept near a reference of zero. So an irrelevant side effect of sipping
soup (slurping) is relevant to the perception of "slobbiness" and it is
controlled.

"Stimulus control" is an irrelevant side effect of controlling a logical
variable because you are not controlling the association between your
responses and the stimuli; "stimulus control" disappears as soon as the
responses required to maintain the logical variable in its reference state
are no longer associated with the stimuli. If you did start trying to control
for "stimulus control" then stimulus control would no longer be an irrelevant
side effect of controlling. But as soon as you started to control for
"stimulus control" (an association between stimuli and responses) you
would lose control of the logical variable.

I think it is extremely important to understand the notion of "irrelevant
side effect" of controlling since virtually all the observations (save a few,
such as the observation of control of interpersonal distance) made by
conventional social science researchers fall into this category.

Kent McClelland (950327) --

Rick's view of what's sociologically interesting probably wouldn't have much
appeal for ordinary sociologists, who, I would say, are aware of space-
invader research, etc., but don't see it as particularly significant. They
may well see it as an "irrelevant side-effect" of something they consider
more interesting, perhaps cultural patterns or "the social construction of
reality".

I agree with this completely though I would substitute the word "legitimate"
for "interesting" throughout. What you say here about sociologists can also
be said for psychologists. Both psychologists and sociologists study
appearances that they find significant (for whatever reason). I am saying
that these appearances are mainly irrelvant side effects of controlling. So
no matter how significant these appearances may seem, they have no obvious
relationship to controlling. It's not that these appearances are not
"interesting"; "stimulus control" is pretty interesting to watch. It's just
that they are not legitimate examples of control; they are, at best,
irrelevant side effects thereof.

I am arguing (as always) that we must determine that we are dealing with
control (fact) before we start to apply PCT (theory). Applying PCT to
existing findings in the social sciences is a truly superficial exercise. It
is like applying PCT to findings in physics and chemisty. For example,
physicists have found that balls accelerate downward at 32 ft/sec^2. I'm sure
that we could develop a control model of a ball that accelerates itself
downward at this rate. Obviously, this would be a ridiculous exercise because
we know (and can show) that balls do not control anything -- acceleration is
neither a controlled variable nor is it a side effect of controlling. Yet we
are doing something just like applying the control model to the downward
acceleration of a ball when we apply PCT to the observations made by
conventional social scientists.

The only _legitimate_ application of control theory is to control phenomena.
Since conventional social science research provides no evidence regarding
what is being controlled -- or that controlling is even going on -- it can
be ignored. Before we start applying PCT all over the place we should have
clear examples of control phenomena to apply it to.

Those who argue that PCT must account for the findings of conventional social
science research should think very carefully about why PCT should _not_ also
be required to account for the findings of physics and chemistry (if you
think it shouldn't). In other words, what evidence is there in the findings
of conventional social science (other than the fact that the results of this
research are produced by living systems) that control is going on?

Best

Rick