Gene: You've stepped into the middle of what seems (IMHO) to be an
almost-religious controversy. One school of thought extolls the virtues
of the causal loop diagram as the entre to thinking systemically about a
problem. The other decries the use of the CLD until AFTER you have built
the stock and flow model ... then the CLD becomes a useful way to
summarize the system in question.
Personally I find myself squarely in both camps ... the argument almost
sounds like a MAC vs PC or WordPerfect vs Word argument (or Wordstar vs
EMACS, depending on where in the baby-boom you find yourself) -- whatever
you grew up on is the best way to look at the world. In some ways, the
discussion about the merits of Vensim vs IThink vs PowerSim is right here.
I grew up in a STELLA/IThink environment, so the S/F approach is the best
FOR ME IN MOST CIRCUMSTANCES. But this seems more than just a matter of
preference ... Barry Richmond, for example, in a very insightful paper on
the HPS web site, suggests that using the CLD as an exploration tool
actually works AGAINST developing a profound understanding of the system
in question, and feels that the use of the archetype by those not already
versed in ST/SD is dangerous and probably counterproductive.
I was in Boston last week and addressed this question to just about
everyone who would stop to listen. Richmond made a strong case for S/F
first, CLD last. John Sterman is decidedly on the same side. Forrester
certainly didn't start with the CLD, for reasons related to the science.
On the other hand, Senge and most of the folks at Innovation Associates
(and MIT for that matter) see power in the CLD as both simple, and
explanatory --- a good introduction to systems thinking concepts. I have
argued against the casual use of the CLD because it is inherently
undisciplined (as you suggest in your last paragraph) ... it it too easy
to create stocks that seem to be affected by stocks ... a no no in
Forrester's eyes.
I take issue with Bob Eberlein's comment that "... the picture was sort of
irrelevant and what really mattered was the formulas contained in the
elements of the diagram." The picture DOES matter if you are trying to
help others use ST and SD tools to solve problems in situ. A modeler
sitting on the other side of the transom can use any method he or she is
comfortable with ... but the objective (and reality) will be a product,
not learning. If the objective is learning, form takes on great
importance. I took this up with him in Boston, and further critized what
I consider to be the worst of all worlds ... mixing the CLD and S/F forms
in the same diagram. That's just too much for me.
Personally, I find the CLD to be easier for people to grasp, but also much
less effective in promoting structural learning. Stocks and flows contain
a much needed discipline --- what are nouns, what are verbs, what things
are conserved, what things are transformed, etc. Once people get really
good at it, anything will do. The question in my mind becomes "what works
best given my current objective." If it looks like we will be going to
modeling, I'd rather start with stocks and flows. If it seems as if we
will be able to gain the necessary insight without resorting to modeling,
the CLD is easier. Trick is to know in advance, I guess.
> The end result being that I am now quite comfortable modeling and simulating
> ST diagrams, and I don't even miss the Stock-Flow diagrams. Now I no longer
> need two different pictures.
That's 'cause you already understand what you're doing. A court case in
England put the game of darts on trial ... if it was a game of chance,
gambling about the outcomes would be illegal ... if it was a game of
skill, gambling would be ok. The counsel called one witness, who picked
up three twenty-penny nails and threw three bulls. Luck or skill? Moral
is, when you're good at it, you can use any tool to get the end you want.
It's novices who buy the $500 titanium adjustable weight and balance
darts.
> Your first thought about this is probably related to how one tells the
> difference between Stocks, Flows, Converters, and Information Flows. I have
> found that color coding the components of the model works just fine.
This is how YOU impose the discipline of S/F diagrams to the CLD. I rest
my case.
Hope this didn't sound too pedagogical ... but it's an area I'm very
concerned about. You only get to introduce systems thinking once in an
organization (innocculation phenomenon), so you have to choose the right
track ... that's where I am right now.
Steve
-- Stephen B. Wehrenberg, Ph.D. Chief, Forecasts and Systems, US Coast Guard; Administrative Sciences Program, The George Washington University; wstephen@erols.comLearning-org -- An Internet Dialog on Learning Organizations For info: <rkarash@karash.com> -or- <http://world.std.com/~lo/>