[From Bill Powers (930826.0700 MDT)]
Michael Fehling, Rick Marken, Oded Maler (930825) --
If the ripples from all this culture shock have died down enough,
there are some interesting points up for discussion.
Somehow the back-and-forth has jogged loose a few brain cells,
which came up with a thought. I suddenly saw an analogy between
the way members of an organization think about the organization
and the way we modelers in PCT think about certain block diagrams
and the equations that describe them. The block diagrams are
meant to remind us of a shared (more or less) perception of how a
certain kind of organization works, a living control system. They
are a model of something. What makes the model work is not, of
course, the equations or the block diagrams, but the physical
components they represent.
In the case of a social organization, the components are the
physical things like buildings, machinery, and articles of
incorporation, which by themselves have no behavioral properties,
and the individual human beings who occupy the buildings, operate
the machinery, and interpret the articles of incorporation (etc).
These individuals have a great many discussions with each other,
one of the products of these discussions being a _model of the
organization_. This model takes many forms. One of them is the
organizational chart, showing who is responsible for what, who
takes orders from and gives orders to whom. Others are a business
plan, a marketing plan, a competitive plan, and other stuff (I'm
pretty ignorant about business so I hope those who know more are
translating appropriately). All the people in the organization
try to behave in such a way as to fit this model, as they
individually understand it and their place in it. There are
problems with that shared understanding, but they aren't my main
point here (I think).
All these models must contain models of the world with which the
organization will interact. This includes a model of the people
-- workers, managers, and customers -- that populate this world.
In fact, the economic aspects of business plans depend pretty
heavily on the model of an individual human being; for example,
are people "maximizers," so they will always respond to suitable
incentives by increasing consumption (or production), or
"controllers," so they are individually aiming for a specific
level of consumption? But that still isn't my main point.
One main point is in the question, "How do people design an
organization if they want to model it as a control system?" The
most immediate answer is, of course, that they have to understand
control systems. As very few people understand control systems,
particularly in the world of commerce, it is more likely that
they will design the organization as a traditional cause-effect,
input-output, stimulus-response system. Of course they have
objectives, goals, and so forth, but when it comes to setting up
a way of reaching them, they fall back on the traditional
understanding of behavior. This understanding says that you must
analyze the external world, construct a plan of action based on
predictions of the effects it will have in the world, and execute
it.
Unfortunately, this understanding doesn't match the way the world
works, or the way people work. The cause-effect model assumes
that the properties of the world will not change, and that there
will be no unanticipated disturbances to throw the projected
future off course. The more detailed the analysis and the more
elaborate and contingency-ridden the plan, the less likely it is
to apply for more than a short time. Business projections and
plans are heavily statistical; by their very nature, therefore,
they can't handle specific events such as a competitor
unexpectedly marketing a more appealing or useful widget than
yours, cheaper.
An organization truly designed as a control system would rely
very little on plans of action, particularly long-range plans of
action. Instead, it would focus on developing the means for
controlling certain variables in present time. This means
learning how to act directly on those variables when they deviate
from the condition one wants to perceive. If disturbances are
reliably predictable -- for example, as a heating-oil supplier
can predict the alternation of summer and winter -- some
proactive actions can be taken. But it's more important to be
able to handle current fluctuations in the controlled variables
without having to know what is causing them. So planning becomes
secondary to understanding the properties of the local world and
setting up control loops accordingly. The important thing is not
to _plan_ actions, but to know how to _vary_ them as required.
I'll let that thought lie there to let others get a word in
edgewise.
What seems to be my main point is that an organization is a model
in the minds of at least some of the people in the organization.
To implement this model, many individuals agree to take on
certain control tasks. Once those tasks are defined (by each
individual), they boil down to keeping certain perceptual
variables in specified reference conditions. The reference
conditions are set, and varied, in part by others higher in the
block diagram of the organizational model, and in part by the
individual's understanding of the requirements of the
organization. They are also highly influenced by other goals that
the individual has, many of which have nothing to do with the
organizational model. When the organizational model is translated
into a working implementation, we have nothing but a set of
individual control systems operating independently of each other,
each according to a private understanding of the situation, and
interacting with each other and the outside world in the manner
in which independent control systems interact.
Well, that didn't come out as crisp as the idea with which I
awoke this morning, but such ideas seldom turn out quite as
wonderful as they seem to the half-awake mind. What do you-all
think?
···
--------------------------------------------------------------
Best to all,
Bill P.