Barry Richard and STELLA

[from Gary Cziko 920910.1558]

I just had my first encounter with STELLA (version 2.1) and it sure looks
like an easy way to do PCT-type modelling. Yes, Rick Marken would go
bananas with this program (sorry, I should've said "MORE bananas")!

But I must say that I was distracted by the manual. I didn't expect to get
a tutorial on control theory, but that's part of what's in the first

On page 18:

"This ongoing, circular control process in which "CONDITIONS LEAD TO
This closed-loop structure is well-known in engineering, the physical and
life sciences, and in medicine. It is somewhat less-established in the
social sciences, although it is every bit as applicable there. FEEDBACK
CONTROL is an extremely important process; one which is absolutely
essential to maintaining viability in a changing environment. We will
revisit this concept several times throughout the remainder of the Guide."

On page page 37 and 38:

"The "internal feedback structure" viewpoint carries with it two additional
perceptual reorientations which also are alien to many people.

"The first is a shift from a "discrete" to a "continuous view of underlying
processes. Feedback loops, and goal-seeking activities, are CONTINUOUS in
their operation. Thus, for example, you do not continue to ship product
flat out until you have ZERO inventory--at which point you ship zero.
Rather, as inventory falls below DESIRED levels, you would begin to
restrict shipments (allowing your "lead time" to rise). Restrictions would
become increasingly severe as inventory fell increasingly far below desired
levels. A feedback viewpoint thus implies "continuous regulation", not
flip-flop! Your STELLA models should reflect this viewpoint.

"The second reorientation in viewpoint, not unrelated to the first, is a
shift from a static/equilibrium perspective to a dynamic/disequilibrium
view. Much of our thinking is static. For example, it is not uncommon to
hear peopl express sentimentslike: "overwork causes stress". IN such
expressions, causality runs only ONE-WAY: from overwork to stress. But in
fact, it is just as legitimate to say that "stress causes overwork". If
you're stressed, you're [sic] mind's not at ease. Maybe you're not
sleeping too well either. Your diet (and exercise routines) may also be
disrupted. All of these influences comgine to result in a decrease in
yhour work-related productivity. Lower productivity, in turn, means less
work gets done, and more errors get created in the process. Slower work
rates, and more rework , mean more work to do. More work means less sleep,
more pressure, more fatigue, and still lower productivity--thereby
reinitiating the cycle. We've all been there. And, the causality is NOT
linear, it's CIRCULAR!"

The author of the manual and president of High Performance Systems (which
publishes STELLA) is Barry Richard, Assistant Professor at Dartmouth (at
least in 1988 when the manual was published). Does anyone know what his
field is? He certainly would fit into the company on this network. Maybe
we give him a free one-year membership in CSG in return for a steep
discount on STELLA for CSG members!--Gary

P.S. Bill [Powers], I am dreaming of playing with the reorganization model
you showed me in Durango on STELLA. But I still have a lot to learn. I
tried to build a simple control system and got the message "no circular
connections allowed" as I put in the last arrow! I think I forgot to add
an "inflow" source to the model (STELLA makes a distinction between "input"
and "inflow" which is not yet clear to me).


Gary A. Cziko Telephone: (217) 333-8527
Educational Psychology FAX: (217) 333-5847
University of Illinois E-mail:
1310 S. Sixth Street Radio: N9MJZ
210 Education Building
Champaign, Illinois 61820-6990

i believe it's actually barry richmond.
he's a system dynamics modeler/programmer. i'm not sure if he came
out of the MIT group or not. as to particular specialty, i really
couldn't say. he doesn't seem to be referenced in any of my system
dynamics texts. what i can point out is that a lot of system dynamics
folks (myself included), while they may have particular areas of concern
and might therefore do a lot of publishing on a core topic (population
studies, economics, urban studies, ecology, etc.), their paradigm causes
them to see the world in terms of systems whose structures are essentially
similar. the modeling method lends itself to viewing apparently different
phenomena as fundamentally related. feedback is feedback, wherever it shows
as for the 'circular connections' error, it's a classic opening mistake.
here's the canonical system dynamics structure (notice that it's essentially
the same as what i showed before):

                /\ : /\
                : : :
                : : :
                : : :
             rate of : rate of

in which the only material flow is that into and out of the level variable.
the rate of flow is controlled by equations which must, in some way, take into
account information about that level. often, there is a pretty circuitous
path, entailing many transformations and inclusion of extra variables.
your 'circular' error arises from violating one of a very few basic rules
of system building. i don't have forrester's "principles of systems" sitting
at my desk right now, but here are three off the top that should help:
1) as just mentioned... any rate that controls flow for a level must
   in some way receive information about that level. this is the fundamental
   assumption of feedback.
2) no rate can control another rate except through its action on a level.
   this one can actually be physically violated both in dynamo and stella,
   and so requires a little care on the modeler's part. one of the skills
   of system dynamics modeling is determining what element of a system to
   actually quantify and call a level. if you can't build the system
   without violating this structural rule, then you've got a clue that
   perhaps you should rethink your approach toward its representation.
3) (here it is...) structural loops cannot be built that would result in
   simultaneously dependent equations. remember that the method involves
   integration in time by intervals. there *is* a calculation order.
   loops must involve something that will delay the information flow.
   instantaneous infinite feedback is disallowed. in dynamo, one can
   actually put a delay equation in the loop, and it will work. in stella,
   the compiler will not even allow you to do that, adhering to forrester's
   original dictum that any loop must contain a level.

-----------< Cognitive Dissonance is a 20th Century Art Form >-----------
Eric Harnden (Ronin)
The American University Physics Dept.
4400 Mass. Ave. NW, Washington, DC, 20016-8058
(202) 885-2748
---------------------< Join the Cognitive Dissidents >-------------------

[from Gary Cziko 92091.1310]

Eric Harnden (920911)

i believe it's actually barry richmond.

You're right, it's Barry Richmond (and NOT Barry Richard) who is the man
behind STELLA. It is also NOT George Richardson (whom Bill Powers
mentioned in his last note.



Gary A. Cziko Telephone: (217) 333-8527
Educational Psychology FAX: (217) 333-5847
University of Illinois E-mail:
1310 S. Sixth Street Radio: N9MJZ
210 Education Building
Champaign, Illinois 61820-6990