Phenomenon of control

peters_r_flc_vaxf.co · March 9, 1992, 2:11am

[From Bill Powers (920308.1800)]

Martin Taylor (920308.1643) --

I take control to be a theoretical construct, not an observation. If it
had been an observation, would it have taken Bill's insight to see it?

It's an observation, but an insight was needed to bring it to attention.

This is the situation observed:

The observed outcome is maintained stable against the disturbance by
variations in the action of the system. The independent disturbance
itself is NOT sensed by the behaving system. The effect of the
transformed action on the observed outcome is nearly equal and opposite to
the effect of the varying independent disturbance on the same outcome.

When the above situation is observed, the behaving system is said to be
controlling the observed outcome and the observed outcome is then called a
controlled variable. The above diagram is an operational definition of
control.

The reason this phenomenon has not been recognized is that "Scientific
Method" (as taught in psychology) systematically rejects controlled
variables. The action of the behaving system (interpreted as a response)
will show a negative correlation with the varying independent disturbance
(misinterpreted as a stimulus), to the extent permitted by the
environmental transformations. A mistaken inference is involved (new page
here):

The negative correlation mentioned above will be maximized only when both
action and disturbance are translated into units of effect on the observed
outcome. If other measures of the action and the disturbance are used, the
obtained correlation between measures of action and disturbance may be
either negative or positive (depending on the measurement scale); it is
also is likely to be considerably less than perfect.

The controlled outcome will show a low correlation with both the actions of
the system and the independent disturbance. The better the control, the
lower the correlation. Therefore the observed outcome or controlled
variable will be discarded by a statistical analysis because it shows no
significant relationship to either "stimuli" or "responses." It will not be
recognized as an outcome of behavior. The mistaken inference will be
that the varying disturbance is being perceived, because it correlates
with a response (and thus satisfies the behaviorist's operational
definition of a stimulus). The slanted line in the second diagram shows
the mistaken inference. The correct inference (shown by the dashed line at the
top of the second diagram) is that the behaving system is sensing
the controlled outcome. It can be verified that when this sensing is
prevented, control disappears.

Thus control is not a construct, but an observation. To detect control, it
isn't necessary to know anything about what is inside the box labelled
"Behaving system." It isn't necessary to know what a control system is or
how it works in order to know that control is going on. The Test for the
Controlled Variable can reveal control by identifying a variable that fits
in the position of the "Observed Outcome" above: The Test is simply a way
of verifying that the conditions shown in the diagram are present.

Best,

Bill P.