Organizations and Organisms

[Avery.Andrews 930831.1159]

I agree with Bill Powers and others' points to the effect that
organizations will be rather different than organisms, whether or not
it is appropriate to regard them as control systems. Rather than argue
abstractly, here's an attempt to lay out some aspects of a hypothetical
case, wherein I stop showing up to give my lectures. I think
things would unfold somewhat like this:

  1st skipped: students notify a secretary or other staff member, who
    assumes I forgot, and reminds me.

  2nd skipped: someone notifies HoD, who gets in touch with me, & tries
    to get some explanation of what is going on, with implicit threat
    that I better start appearing.

  3rd skipped: HoD talks to Dean, who issues serious threats.

  ...

  nth skipped: Dean notifies Vice Chancellor, who issues even more
    serious threats

  ...

  n+mth skipped: offender sacked, & new lecturer appointed

In phase 1 (1st lecture skipped), either a secretary or staff member of
the department functions as a (fairly low gain) control system, wherein
a perception of what somebody is doing differs from the reference value
(what they ought to be doing), the error signal causing the person to
chat with the offender about the offense. I conjecture that this
control system exists in similar form all the staff-members plus the
secretaries (people with some kind of responsibility for getting the
business of the department done), but not technicians, cleaning staff,
members of other departments, etc., and that the perception it controls
for is something like `the business of the department goes smoothly'.
  I say that the gain is low because
these people would not go to great lengths to locate the offender: the
matter would only be raised if they encountered them.

If the offense is not repeated (this control system appears to effect
control), that will be the end of it, but if somebody becomes aware that
two lectures have been missed, they would probably notify the HoD (so
far, things remind me a bit of neural recruitment in mammalian nervous
systems). At this point I'd say it was sensible to regard the informer
as a perceptual function for a control system located mostly in the HoD,
concerned with `ensuring that Dept. members perform their duties'.
So we have a set of interchangeable low-level, low-power control systems
that alert a higher-level one when their error-signal persists.

The HoD's system will function at a considerably higher gain than the
previous one - it will get me located, demand an explanation of what
is going on, etc. What happens next depends on the circumstances, but
the invariant is that HoD activity will continue until there is an
appropriate person fronting up for the lectures - myself if I've been
being sick and am better, perhaps a replacement if it looks like I'll
be out of action for a long time. On a shorter time-scale, the HoD
will be controlling for the belief that someone *will* front up for
a lecture pretty soon - e.g. I can fob him off by *saying* that I've
been sick, but will appear at the next lecture. But when I don't show
up for lecture three, HoD activity starts again (given that someone or other
serves as the relevant perceptual function). The HoD might very well
inquire of a student whether I had appeared, e.g. actively recruiting
a perceptual functor.

When the HoD's system fails to achieve control, it recruits a more
potent one, serving an alerting-and-perceptual function for the dean,
and so on--I'm not really sure how things would work at the higher
levels of the administration. The dean might talk to me directly,
or suggest courses of action to the HoD, who would therefore be
providing lower level perceptual and control systems for those in the
Dean. But the dean would definite *not* front up in the lecture hall
at the time I was supposed to appear for my next lecture, tho the HoD
might, in his capacity as perceptual subsystem for the Dean.
Similarly for the Vice Chancellor, recruited by the Dean if the
error signal persists.

On the basis of all this, I see considerable merit in thinking of an
organization as a big control system, although we certainly want to keep
in mind that its pieces are located in different individuals, and that
the `glue' that holds it together is different from what holds nervous
systems together (basically, organzations function by virtue of the
people in them knowing what is supposed to be happening, while nervous
systems don't function by virtue of the neurons in them knowing this).

One thing that strikes me as interesting about all this is that
regardless of the written rules, people in different places in the
organization will feel responsible for quite different aspects of
what is going on in it - if some German student showed up at my door
wondering where his German class was, I wouldn't say anything to the
German secretary about it, and would find it rather odd if she said
anything to me about a linguistics class I missed. But I'd find it
perfectly normal to communicate with the German secretary about, say,
some aspect of the building plant, like heating failure, if our own
were unavailable. So the structure is surely looser and more flexible
than neural ones, but there does seem to be some there (and mammalian
neural systems seem to be a lot looser and more flexible than those
of insects, and maybe neurons are smarter than we think they are).
And, the unwritten rules might be more rigid and binding than we would
like to think they are.

Avery.Andrews@anu.edu.au

From Tom Bourbon [930901.1212]

[Avery.Andrews 930831.1159]

I agree with Bill Powers and others' points to the effect that
organizations will be rather different than organisms, whether or not
it is appropriate to regard them as control systems. Rather than argue
abstractly, here's an attempt to lay out some aspects of a hypothetical
case, wherein I stop showing up to give my lectures. I think
things would unfold somewhat like this:

[Here Avery described the expected series of "escalating" responses from
people in the administrative hierarchy at his institution.]
..

In phase 1 (1st lecture skipped), either a secretary or staff member of
the department functions as a (fairly low gain) control system, wherein
a perception of what somebody is doing differs from the reference value
(what they ought to be doing), the error signal causing the person to
chat with the offender about the offense.

I might describe your 1st-level person (watch person) just a little
differently, as having a reference value of "perceive other people doing
certain things," against which to compare present perceptions of the things
others are doing. Perhaps in this adopted role a watch person *seems* to
have low gain, but for all of the person's activities such as standing,
sitting, speaking, patrolling the hallways, peeking into classrooms, writing
memos, punching buttons on telephones, typing, drinking coffee, maintaining
physiological processes, and the like, the person must be in a high-gain
negative feedback relationship with the world.

I conjecture that this
control system exists in similar form all the staff-members plus the
secretaries (people with some kind of responsibility for getting the
business of the department done), but not technicians, cleaning staff,
members of other departments, etc., and that the perception it controls
for is something like `the business of the department goes smoothly'.

Perhaps my previous comment involved a matter of differences in our ways of
expressing the same idea. Here, I have the impression that we might
interpret the situation differently. Perhaps you can help me get a better
undrstanding of what you mean. When you spoke of the 1st person who would
notice your absence and say something about it, the person was thought of as
a control system with a reference value of "perceive other people doing
certain things." In this section, my understanding is that you are saying
there is a different control system, more than the persons in it, and that
the new system has a reference value of "see the business of the department
go smoothly." It seems to me that this would necessarily be a reference
value in each person, not in a supra-individual 1st-level system.

Let me explain a bit. If each person assigned the role of "watch person"
had a reference of seeing the department work smoothly *and if* one of the
criteria for smooth working is people (like faculty) doing certain things
(like meeting classes), we would have a hierarchical relationship among
reference signals, with a hint of the program-level being part of that
hierarchy. (I am a watch person. The principles by which I maintain my
role as watch person necessarily entail following certain program steps ...
and so on.) Every individual watch person would necessarily "possess" a
similar arrangement of reference values, or a functionally equivalent
alternative arrangement. Let me know, Avery, if I am misreading you, or if
you think my suggestion is similar to or different from what you intended in
your post.

I say that the gain is low because
these people would not go to great lengths to locate the offender: the
matter would only be raised if they encountered them.

Understood, but what if the program does not call for "searching out"
possible offenders? It might call for "saying something to a possible
offender when the opportunity arises." When the watch person spoke to the
alleged or suspected offender, the speaking would be accomplished by a
high-gain system.

If the offense is not repeated (this control system appears to effect
control), that will be the end of it, but if somebody becomes aware that
two lectures have been missed, they would probably notify the HoD (so
far, things remind me a bit of neural recruitment in mammalian nervous
systems). At this point I'd say it was sensible to regard the informer
as a perceptual function for a control system located mostly in the HoD,
concerned with `ensuring that Dept. members perform their duties'.

Understood, but what is gained, and what is lost, by referring to the
informant as a "perceptual function," meaning by that phrase that the
informant is logically and functionally equivalent to a perceptual function
in a formal PCT model? I am attempting to come to a better understanding of
why so many people are interested in drawing this analogy, or in going
beyond analogy to assert that such an informant *is* a perceptual function
in a social control system. I am not trying to characterize your remarks
as good or bad or anything else; I am simply asking if you can help me
understand this point.

Does the informant notify a control system located in the H0D, or
alternatively does she or he notify the H0D, whom we might want to model as
a control system, some of whose reference values pertain to perceptions
different from those of the watch person? (Do you think at least some
Heads of Departments would also have, somewhere up in the hierarchy, a
reference value of "see the business of the department go smoothly"?)

So we have a set of interchangeable low-level, low-power control systems
that alert a higher-level one when their error-signal persists.

By this, do you mean that, at any particular time, any one of the watch
persons could be the one to inform to the HoD? If so, what is gained or
lost (I am willing to see suggestions in either direction, or both) by using
the PCT-like language? Could you try to illustrate this organizational
control system in diagram along the lines of Figure 4 in my recent post? If
you could, it might help me have a clearer understanding of what you are
saying.

The HoD's system will function at a considerably higher gain than the
previous one - it will get me located, demand an explanation of what
is going on, etc.

Again, why would we think the HoD has a higher gain? Could it be that the
HoD has a different reference, perhaps one for "locate possible violators"
or some other equivalent? Whether or not a person searches out possible
offenders seems to me to be more a matter of different reference values,
than of different gains. Either person (seeker and non-seeker) would still
operate as a high-gain negative feedback control system. Does that sound
like a defensible alternative to the ideas in your post?

I do not presume to have the final and correct resolution of the question
about whether organizations *are* PCT systems, but I believe any answer must
preserve the understandings we are working out concerning individuals. In
this instance I mean that in our quest to better understand how to
conceive of individuals in social organizations, we ought not, without good
reason, casually abandon a seemingly distinctive and foundational assumption
of PCT: living control systems exist in high-gain negative feedback
relationships with their environments. Until that insight is shown to be
wrong, I take it as a principle with which my ideas about social
interactions and my PCT modeling of interactions must be consistent. If my
modeling of interactions fails, and the failure seems to stem from that
principle, I would first go back to the individual level and look to see if
the notion of high-gain negative feedback is really warranted.

[Here, Avery described subsequent events as knowledge of his absence from
class worked its way up the administrative hierarchy. My comments on these
paragraphs would be similar in most details to what I have said about
the first two levels.]
..

On the basis of all this, I see considerable merit in thinking of an
organization as a big control system, although we certainly want to keep
in mind that its pieces are located in different individuals,

As you probably have guessed, I do not immediately see merit, considerable
of otherwise, in thinking about the organization as a big control system. I
do not say that to denigrate your position or your ideas; I merely do not
see it. For example, I do not understand the meaning of the statement, "its
pieces are located in different individuals." When I try to imagine an
organization as a control system model, I keep seeing the reverse of what
you say; on the organization chart, the individuals are located in different
pieces of the chart. Then, in a manner described effectively by Bill Powers
(I do not have the posting date readily at hand), people "in" different
boxes adopt reference values they believe will make them good watch persons,
or HoDs or whatever the chart says they are supposed to be. Please help me
try to see this the other way; many people seem to share your view of this
reversible figure and I want to understand it.

My post with figures was part of *my* effort to comprehend the ideas you
and a number of others express. It would help me if you were to
present your own version of my Figure 4. In it, I showed each person
playing the part of a box in an organizational chart; show me how the boxes
in the charts would, instead, be in each individual. Perhaps we will
discover that our ideas are not that far apart at all.

[continuing the sentence interrupted above] ... and that
the `glue' that holds it together is different from what holds nervous
systems together (basically, organzations function by virtue of the
people in them knowing what is supposed to be happening, while nervous
systems don't function by virtue of the neurons in them knowing this).

I agree with you that this is a very different kind of "glue," but I have
not been able to bring myself to the point of thinking difference in the
glue can be ignored when we move from cells to individuals to organizations.
An individual who controls knows what he or she intends to perceive; each
individual in an organization knows what she or he intends to perceive; I
am still not convinced that while I type this reply, a muscle cell in my
finger, or a cone in my retina, know what "I" intend to perceive, or that
they have references for perceptions. Those cells can be represented as
input-output devices. (Rick Marken has posted on this theme many times.
There is little I could add to what he has said.)

In my PCT modeling of interactions, the "glue" that holds two people
together (or two models, or two hands on one person, or a person interacting
with a PCT model -- my usual song and dance) -- that "glue" seems to be
the linkages between actions by individuals and the effects of those actions
on variables in a shared environment. If two or more controllers affect
variables in common, then the actions of the controllers become coordinated.
The nature of the coordination depends entirely on the references of each
controller: given the exact same linkages between actions and affected
environmental variables, two controllers can, by their individual selections
of reference perceptions, turn the setting into one of interference
(unilateral or mutual), of cooperation, of pure conflict, of one controller
aiding or assisting the other, of one controller (A) controlling the actions
of the other (B), of B (whose actions are controlled by A) producing
counter-control of the actions of A, and so on. The interactions can
become quite "abstract." For example, in the event that one system aids the
other, the aided system can set and control for a reference perception of
"controlling the actions of the aiding system." Living control systems and
PCT models of control systems can create a remarkable fabric of
interactions out of exactly the same physical linkages between just a few
simple variables (the positions of two handles, one cursor and one target
are all it takes.) My awareness of that great array of social
interactions, all created by simply changing the reference perception of one
or both of the controllers, interferes with my grasping the need for, or
seeing the merit of, thinking of social interactions as occurring in social
control systems, where each person *is* a different function in the system.
I guess I am saying, "show me," either in a working model, or (with much
less effort) in a diagram of the proposed social system. Then maybe I can
see the other side more clearly.

Until later,
Tom Bourbon
Department of Neurosurgry
University of Texas Medical School-Houston Phone: 713-792-5760
6431 Fannin, Suite 7.138 Fax: 713-794-5084
Houston, TX 77030 USA tbourbon@heart.med.uth.tmc.edu

[Martin Taylor 930903 11:30]
(Tom Bourbon, Avery Andrews, Michael Fehling various postings)

A year or so ago (my term for 3-36 months) we had a somewhat inconclusive
discussion about distributed control within a single hierarchy. It seems
to me that much of that discussion can be (and perhaps is being) replayed
in talking about organizations. Here's a brief (I hope) summary.

I probably won't see any responses to this until October. I expect it
to be controversial, and will be interested to follow any discussion on
it when I come back.

When we draw or model specific control hierarchies, we allocate one ECS to
each controlled perceptual degree of freedom, and sometimes that allocation
is supported by argument that reorganization will bring ECSs to a state of
near-orthogonality. But we know that real nervous systems rely on what one
might call "coarse coding" (following the neural network people). When we
move a joint in a particular direction, many different control systems in
the motor cortex (or at least many outputs from neurons there) work in
parallel, each being optimally tuned to a slightly different direction, but
the whole group summing to the direction intended. By analogy, imagine a
"rubber band demo" with hundreds of bands tied to the knot in the middle.
If a disturbing variable tries to move the knot eastward, it will not only
be the west-pulling band that applies the correction, but also many of the
bands pulling nearly west, perhaps as far away as northwest and southwest.
Together, they provide the control that we tend to model as "controlling
for westwardness." We probably would model the whole system as having
two ECSs, one for east-westwardness, and one for north-southwardness. And
the model would probably fit very well.

In the earlier discussion, I tried to extend this idea to higher levels,
arguing that what an observer might see as a single control of some perceptual
variable apparent to the observer could actually be performed as a whole
lot of parallel controls in which the observer's variable did not appear
as such. Only its component elements might appear as perceptions within
the controlling hierarchy. Any apparent coordination, and effective control
of what the observer thinks to be a CEV, would occur because the individual
components of the supposed CEV were individually controlled, not because
the observed hierarchy had an ECS for that CEV.

(Parenthetically, this suggests another component of the Test: the tester must
succeed in changing the value of a lower-level CEV, and determine whether
the proposed CEV that has it as an element is kept under control. Simply
observing a proposed CEV at its own level is not enough.)

The point of this discussion is to suggest that a theorist's ability to
model a control hierarchy with a small number of appropriately connected
ECSs is not a demonstration that the real hierarchy is constructed in the
same way. As Rick has pointed out, modellers have satisfactorily modelled
human behaviour in some circumstances by using a single ECS, even though
we know full well that the human is more than that, even in a simple
tracking task.

What Michael seems to be doing is to try to model the observable behaviour
of an organization AS IF it were a hierarchic control system, without
needing to assert that it works by having trans-personal ECSs that
receive percepts from and provide references to ECSs within individual
persons. Just as in distributed control inside an individual there nood
not be an ECS that controls a CEV that an observer sees to be controlled,
so within the organization, the controls performed by individuals can add
up to the appearance of a CEV being controlled by the organization, even
though the internal structure of the organization has no corresponding ECS.

We "know" the organization has no transpersonal ECS only because we can see
inside the organism's structure, and we can find no corresponding entity.
We can see very little inside an organism's nervous system, so we are not
constrained by the fact that we have not seen *the* entity corresponding
to a particular ECS we model.

If Michael's models work to describe the observable behaviour of an
organization, would they not have as much validity as the models many
of you use to describe the behaviour of a person? In both cases, we
could say that the model is known not to describe the structure. It
describes the way the structure works, and that is different.

Martin