[From Bill Powers (980913.0358 MDT)]
Bob Eberlein --
I have some questions about your model wfinv3.mdl. Is this just an example
of the type of system design, or does it represent realistic behavior? The
reason I ask is that the oscillations don't appear very useful; in fact
they cause the inventory to rise and fall repeatedly, while the size of the
workforce does the same. If you were making recommendations to a company,
is this what would your design look like?
Best,
Bill P.
[From Bob Eberlein (19980914.0640 EDT)]
Hi Bill,
I was a little bit baffled by the responses that you and Richard sent to my
last post. I think this question may help to explain a little of my
bewilderment (though only a little).
The inventory workforce model is a classic in system dynamics. Jay Forrester
was doing work for Sprague Electronics and they were complaining that the
seasonal fluctuations in their demand were somewhat unpredictable and this was
wreaking havoc with production scheduling. After digging in to the problem Jay
discovered that there were no seasonal fluctuations in demand. The company
itself was producing these oscillations in production because of their hiring
and inventory replenishment policies.
In short, this model is meant to be a representation of how people operate -
and most system dynamic models have that character. The fact that they operate
pretty poorly in many settings is troublesome to many, but a fact of life
nonetheless.
So no, I am not trying to design any any sort of control system here, just show
stuff that might be going on. That is also the approach I took with the Speak
model.
Bob Eberlein
[From Bill Powers (980913.0358 MDT)]
I have some questions about your model wfinv3.mdl. Is this just an example
of the type of system design, or does it represent realistic behavior? The
reason I ask is that the oscillations don't appear very useful; in fact
they cause the inventory to rise and fall repeatedly, while the size of the
workforce does the same. If you were making recommendations to a company,
is this what would your design look like?
[From Bill Powers (980913.0758 MDT)]
Bob Eberlein (19980914.0640 EDT)--
I was a little bit baffled by the responses that you and Richard sent to my
last post. I think this question may help to explain a little of my
bewilderment (though only a little).The inventory workforce model is a classic in system dynamics. Jay Forrester
was doing work for Sprague Electronics and they were complaining that the
seasonal fluctuations in their demand were somewhat unpredictable and this
was
wreaking havoc with production scheduling. After digging in to the
problem Jay
discovered that there were no seasonal fluctuations in demand. The company
itself was producing these oscillations in production because of their hiring
and inventory replenishment policies.
Beautiful! So by playing with the parameters of this model, Forester was
able to make it show the same "seasonal fluctuations," only without any
fluctuations in demand? I trust that he then showed Sprague how to adjust
the "time to adjust workforce" and introduce a little damping to eliminate
the oscillations.
In short, this model is meant to be a representation of how people operate -
and most system dynamic models have that character. The fact that they
operate
pretty poorly in many settings is troublesome to many, but a fact of life
nonetheless.
I think it's not so much the way people operate as the way the systems they
invent operate -- the policies, rules, customs, organizations, and
understandings. Where cognitive understanding and conscious choice are not
so dominant, people operate in much more predictable and close-to-optimal
ways.
Also, in the realms where system dynamics illuminates relationships, there
is no particular kind of system that we can expect to find. Any
relationships people can imagine can be set up, and operate through
interaction with natural laws. So you can have stimulus-response systems,
positive feedback, negative feedback, and passive equilibrium systems. You
can have adaptive systems and systems that resist adaptive change. Of
course certain common system components, like stock-and-flow elements, may
often be seen, but they can be hooked up together any way people think
useful (or don't think about at all, but accept because somebody else did
it that way).
The focus in PCT is very different from that of SD, although it's
complementary. What PCT is concerned with is how ONE INDIVIDUAL works. How
can a person keep a finger on a randomly-moving target? How can a person
stand up and balance, and walk? How can a person decide to hire someone,
and then do all the things necessary to make that happen, even when
difficulties and distractions occur?
The things PCT is concerned with show up everywhere within the SD models
that human actions are required to make something happen. A cognitive
system can come up with the decision, "Hire more machinists." But that is
only a verbal description of the desired outcome, that more machinists
become hired. Given that objective, the personnel worker must then figure
out what to do to realize that desired end -- to create a perception
reporting that more machinists have, indeed, been hired.
The proper model for this sort of action is not a stimulus-reponse model,
but the hierarchical control-system model of PCT. PCT is the theory of
purposive behavior. In the real world, you don't get the same result as
before by repeating the same actions as before. Systems that work in the
real world do so by specifying an outcome, then acting continuously on the
environment to reduce the difference between what _is_ being perceived and
what is _intended to be perceived_. Real living systems operate to achieve
repeatable ends by variable means, and they must do so because the
environment is continually changing. Systems that can do this are called
negative feedback control systems; they are the building-block of the PCT
model of human nature.
Any place in an SD model where a person is called upon to create a
consistent or predictable result, PCT is needed to explain how doing that
in a changing environment is possible. The question of whether that result
is advisable or even exactly the wrong thing to do, in terms of larger
goals, is the business of SD. But to explain how goals can be set and
achieved, aside from their advisability, is the business of PCT.
That, in a nutshell, is what PCT and SD have to do with each other.
Pretty soon, I'd like to send you either a program for taking tracking data
or the data from such a program. If you're willing, I'd like you to show us
how to use Vensim to fit a control-system model to the data and come up
with values for the model's parameters. I have a hidden agenda, of course:
I hope that when you do this, you will be intrigued by the way the control
model explains the data, and get hooked on PCT as an adjunct to SD. The
first sample, naturally, is free.
Best,
Bill P.