REPLY Definition of root cause (SD6829)

[SDMAIL Jack Harich]

Posted by  Jack Harich <register at thwink.org>


SDMAIL Jean-Jacques Laublé wrote:
> Hi Jack.
>
> This is a very theoretical discussion, and I try to apply SD to very 
> concrete problems. I think that to illustrate your ideas and to prove 
> that they can have a practical application, it would be a good idea to 
> show an example with even a very simple with a model (snip)

Be glad to. Just look at the Dueling Loops model. It has the root cause, 
the low leverage points, the high leverage points, and solution 
convergence. The full paper version is the better read, due to more 
detail.  This example uses SD. See:

http://www.thwink.org/sustain/articles/005/DuelingLoops_Paper.htm

For a second example see:

http://www.thwink.org/sustain/manuscript2/ProperCouplingPackage.htm

This uses causal flow diagrams instead of SD. This allows the dynamic 
hypothesis to be much more easily understood.


Jean-Jacques Laublé wrote:
> Hi Jack
>
> Can your models be applied  to a precise case and preferably not
> needing years to get the results, so as to prove its usability?
> Is it possible to make the model run on past data to see that it 
> replicates
> past results?
> I need some motivation to start studying your models.

Actually it's not the models that are applied, but the problem solving 
process. The one that works for me is the System Improvement Process 
(SIP). This was designed to solve any difficult social problem. The 
links to the Dueling Loops and Proper Coupling models are merely 
examples on how the process can be applied. SIP is itself an example of 
how the better the process fits the problem, the easier the problem is 
to solve.

"Is it possible to make the model run on past data to see that it 
replicates past results?" - Both links are to examples of concept 
models, not calibrated models. The Proper Coupling model is a causal 
flow model, so it doesn't run physically. But you can run it mentally to 
see that it replicates past system behavior.

The Dueling Loops model is a stock and flow model, so it runs. In the 
long version of the paper 22 simulation runs are discussed. This should 
be more than enough for you to see how the model behaves, given various 
starting states. Note that concept models do not duplicate historic 
results exactly. They only duplicate general modes of system behavior. 
If your problem is how to shift from one system mode to another, and not 
how to fine tune (optimize) a system to be a little better here and 
there, then concept models can suffice.

Hope this helps you get going. One question to ask yourself is what 
process do I use to solve problems with SD? If you are not using a 
formal process, checking off the steps as you go and continuously 
improving it, then it's likely you will stumble around and confuse 
yourself and clients, unless a problem is easy or you have encountered 
it before.

Posted by  Jack Harich <register at thwink.org>
posting date  Thu, 20 Mar 2008 09:30:32 -0400