[From Rick Marken (2003.05.23.1500)]
Marc Abrams (2003.05.23.1501)
PCT control system error and
reference conditions are not the same as system goals and mistakes. Very
often we speak of “error” when we are talking about mistakes, and when
we speak of reference conditions we are really talking about goals. It
gets confusing. These words are not interchangeable, yet we all use them
as if they were.
I agree that error in a control system is not the same as a mistake.
But I think the reference condition in a control system is exactly what
we mean when we talk about people having goals.
I think of a mistake as an outcome that was not intended. When I press
“5” rather than “2” when dialing the phone, that is a mistake if my intention
was to press “2”. Such a mistake may or may not be associated with control
system error. If it is associated with error, then the mistake should be
quickly corrected as the control system acts to bring the result (“5”)
to its goal (reference) state (“2”). If there is no error then the
mistake is not corrected. A mistake is typically defined with respect
to the intentions of the actor. So if the actor’s intent was to press “5”
and “5” was pressed, then there was no mistake even if an observer knows
that “5” was, in fact, the wrong button to press if the actor was calling
his mother, say. In my “prescribing error” paper I suggested that we call
this kind of “mistake” a “failure”. A mistake is, then, an unintended outcome.
A failure is an intended outcome that is judged wrong by an observer (such
as a teacher). It seems to me that it would be very hard to make such distinctions
without the help of PCT.
Regarding “goal” and “reference” here is Webster’s Revised Unabridged
Dictionary definition number 1: the state of affairs that a plan is intended
to achieve and that (when achieved) terminates behavior intended to achieve
This sounds a lot like the reference state of a controlled variable
to me. Since that reference state is specified by the reference signal,
I tend to think of the reference signal (or the value of that signal, the
reference condition) as the goal. But the word “goal” is ambiguous because
it can refer to the reference signal value or to the reference state of
the controlled variable itself. For example, a thermostat is
set to 70 degrees. That’s the reference condition. The actual temperature
(the controlled variable) is whatever it is at the moment. When the temperature
is 70 degrees is in the reference (goal) state. So when we say that the
thermostat’s goal is 70 degrees do we mean that the reference is set to
70 or that a temperature of 70 is the goal? If it’s the former, then the
goal is always there, in the thermostat, as the reference. If it’s the
latter, then the goal doesn’t exist until the temperature in the room (the
state of the controlled variable) is actually 70 degrees. I prefer the
first interpretation but understand the second as legitimate.
This illustrates the problem of mapping the control model to everyday
language. Defining words like “goal” and “mistake” so we know what
we’re talking about is certainly a good idea. But I think that the clearest
definition of what we are talking about is given by the model. When
there is ambiguity, we can always go back and say what we mean in terms
of the functions and variables in the model. I don’t think it is
possible to develop a lexicon or dictionary that is as clear and unambiguous
as the model itself.
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971