PCT, organizations, and orgamisms

From Tom Bourbon [930830.1007]

Last week I repiled to a post form Michael Fehling. My reply was intended
for distribution on CSG-L, but it went to Michael directly. His reply came
directly to me. We agreed that I would try once again to send my post, with
his reply appended, to CSG-L. Here it is.

From Tom Bourbon [930827.1341]

Michael Fehling (939826)

Michael, my first comments will be directed to the end of your post, which I
therefore rearrange.

My thanks to Rick Marken and, more recently Tom Bourbon
(930826.0931) for trying to set me straight on whether the
environmental context is modeled in the PCT methodolgy. In fact,
Tom, I actually did believe that it is. But, a few peple replied
to an earlier comment of mine by claiming that PCT is indeed
"radically subjective." (In my earlier message, I had defined
radical subjectivity as the kind of solopsism one finds in
bayesian decision theory.) So, I got the (evidently false)
impression from these remarks that indeed the only way the
environment is modeled in PCT is in terms of the controller's
perceptions. Having so interpreted these remarks, I didn't want
to be glib in finding communalities between PCT and my research
until this issue was cleared up. You've done that now. However,
now I no longer can interpret those claims that PCT is radically
subjective. Oh well, that can wait.

I am pleased to see that these exchanges on the csg net have begun to "pay
off," in the sense of enhancing your awareness of some of the less obvious
features of PCT modeling. I withheld commenting on the part of your
earlier post in which you said PCT is "radically subjective," preferring
to wait until you gave additional information about your work as a context
for interpreting your statement. As you have discovered, the environment is
very much a part of the PCT model. I can imagine how that fact might *seem*
to contradict our assertions that PCT models are designed from the
perspective of the modeled system, but I believe the contradiction is
illusory. Let me explain the rationale for my claim.

In the PCT model, the person-organism-system is modeled as knowing its own
perceptions of the environment, but not the environment in and of itself.
As observers and theorists, we are in the environment of the controlling
agent, where we perceive both the agent and certain variables that affect and
are affected by it. (Those are *our* perceptions.) When the agent acts so
as to affect its own pereptions, it affects certain variables in its
environment. Simultaneously, other environmental variables (disturbances,
perhaps including results of our actions as observers) affect the variables
affected by ("controlled" by) the agent. That is the state of affairs in
which an agent acts to control its perceptions. (When they control their
perceptions, people probably will say they are controlling various features
of the world. PCT theorists interpret and model these events as the person
controlling perceptions of the environment.)

Within the PCT model, on the other hand, we have adopted a view that
includes both the agent and what in a particular instance we believe to be
relevant environmental variables. Our belief concerning the relevance of
certain variables is subject to falsification via "the test for the
controlled variable," mentioned by several people in replies to your posts.
We identify the variable(s) we believe the agent controls, then we act to
alter the state(s) of the variable(s). If the agent acts to oppose or negate
our effects on the variable(s), then we count that as possible evidence that
the agent controls its perceptions of the suspect variable(s).

Only when we have such evidence in hand do we assume that a particular
person-organism-system is acting as a control system, and only then can we
justify using the model of behavior in PCT to represent the person-organism-

To summarize these ideas as they pertain to the seeming contradiction with
which I began, we model *the agent* as knowing only its own perceptions of
the environment, but *our model* also includes the parts of the world *we*
believe affect and are affected by the agent.



The second point on which I wish to reply has already drawn comment from
Bill Powers (930827.0745 MDT). You said:

Descriptively, we aim to show how an organization and its members
faithfully reflect the constitutive features of a control system
and embody control functions. Although relatively few social
theorists have attempted to describe organizations in this way,
the jobs turns out to be surprisingly easy. When one looks
closely at any real world organization one finds it constantly
adapting to disturbances. The ones that don't look this way at
first turn out to be merely operating in very stable conditions.
... Structurally, it is easy to
find organizational members playing the role of input transducers
by monitoring the environment and gathering data, others carrying
out perceptual transformation as they analyze and interpret these
inputs, others defining the reference perceptions to which the
transformed inputs are compared, for example, by (re)setting
performance objectives, and still others whose role it is to
produce the outputs as a function of the comparison process, etc.

As Bill indicated in his post, the approach you described *appears* to run in
the opposite direction from the one we have adopted in our work, ("Our
work" referring to the work of those who attempt to use PCT to develop
working models of control systems.) Our first step is to identify what we
think might be a controlled variable. Having found one, we next look for a
likely candidate for the role of controller. If one appears, we proceed to
"the test." Bill has described the steps and events that follow from
there. If it succeeds, our activity culminates in the development of a
functioning PCT model of the target system -- a model that, when it is run in
simulation, produces results of its behavior that very closely resemble
results produced by the original system. This successful modeling
constitutes what might be called "the second test" in PCT, the first being
the identification of a controlled variable.

I could say more on this point, and on other sections of your post where I
believe it appears, but I prefer to wait until you reply.


Bill Powers Bill Powers 930826.0800 MDT) argued to Avery Andrewss
that "there is an essential difference between the way components
inside an organism relate to each other and the way whole
organisms relate to other whole organisms -- particularly in
social organizations." As usual, Bill made some very insightful
comments that support this conclusion. As it turns out, I think
that Bill's arguement is dead wrong. But that is not the point I
want to make. Instead, I suggest that this is a topic that can
easily be deferred until scientific theory (especially PCT and
HPCT) has been developed that help us say more clearly what we
mean by an "organism" and an "organization."

On this point, I am certain there will more for all of us to say in the
future, for I believe most PCT modelers will say Bill's argument was
dead right!

Until later,



Date: Fri, 27 Aug 93 19:40:09 -0700
From: fehling@lis.stanford.edu (Michael Fehling)
Message-Id: <9308280240.AA18570@lis.stanford.edu>
In-Reply-To: "Tom Bourbon"'s message of Fri, 27 Aug 93 13:51:30
Subject: PCT, organizations, and orgamisms


Since you posted your remarks to me alone, I reply in like manner.

  First, in my earlier post I didn't say that PCT _is_ radically_ subjective,
I tried to get Bill or someone else to accept or reject this statement. In
fact what I said to Bill was, "I hope that PCT is not 'radically subjective'".
I went on to define a theory as radically subjective if it disallows any
mention of "facts" about which the agent has perceptions or beliefs and,
hence, any discussion of relations between these facts and their corresponding
perceptions. In fact, Bill replied by saying "In a purely theoretical sense,
PCT is indeed radically subjective." So, I had to await further opportunties
to inquire more concretely.In fact, Bill replied by saying "In a purely
theoretical sense, PCT is indeed radically subjective." Others gave a similar
  I had expected Bill and others to simply say something like this:

In PCT one models those environmental variables with respect to which
the agent is exercising perceptual control. WHen an agent acts to
affect its perceptions PCT can also model the way in which this
process of perceptual control affects those environmental variables.
Thus one can ask questions about the relationship, if any between
perceptual control and environment impact.

I suspected that Bill was responding something other than what I had defined
as radical subjectivity. Thus, I sought further opportunities to get an
answer to my question. Rick's message and then yours of a few days ago
satified me that PCT is saying essentially what I had believed (and hoped).
Your private message to me today repeats that point.

  Similarly with "the Test." This seemed clear to me, too. In fact, it is so
clear that I didn't think it worth elaborating in my remarks about why I think
an orgnanization is a (perceptual) control system. But, to be explicit, I
tried to say that, first and foremost, I cannot think of any organization for
which no controlled variables can be found. It is trivial to find them. I
said this, but maybe not clearly enough. So, I moved on to discussing the
structural properties of an organization as a control system--e.g., what
entities in the organization embody the input funtion, the perceptual
transform, etc.

  So, I don't think my approach "run[s] in the opposite direction from [PCT]."
Finding the plausible controlled variables for any organization is a no
brainer. Defining the exact nature of the control functions and their effects
on the environment is what I hope PCT gives better than other approaches to
control. Another similarity is that my work, too, culminates in the
development of a testable simulation model.

  Finally, as to whether Bill is right or wrong on the similarities between
organisms and organizations as control systems, I was urging that both you
PCT'ers and others such as I would be better off putting our efforts into
converting this type of assertion into testable hypotheses rather than
spinning our wheels defending vague opinions. I don't care whether everyone
in the PCT community agrees with Bill or me. I care to get results that,
first, clarify the questions, and then, answer them.

  If you meant to post this exchange to the whole list please feel free to do

- michael -