Subject: Ways of understanding a dynamic system

Posted by Paul Newton on 5/10/2004
In Reply To:Ways of understanding a dynamic system Posted by Lees Stuntz on 5/10/2004

Message:

Lees,


I can think of two "hierarchies" but I'm not sure either of them can reasonably be thought of as hierarchies in the sense of "degrees of understanding dynamic systems," but it might be possible to fashion them this way as I discuss after listing them below. The two I can think of are:


1) the popular "Event - Behavior - Structure" hierarchy,


and


2) Forrester's hierarchy of systems structure as described at the beginning of his Urban Dynamics book (Chapter 2 I think), and in his market growth paper. Here is is from highest level first, to lowest level:


Highest level: Closed boundary around the system


Next lower level: Feedback loops as the basic structural element within the boundary


Next lower level (2 items):


- Level (state) variables representing accumulations within the feedback loops


and


- Rate (flow) variables representing activity within the feedback loops.


Lowest level: (4 items that are part of the rate (flow) variables and that represent policy structure components)


- Goal


- Observed condition


- Detection of discrepancy


- Action based on discrepancy


(End of # 2 )
----------------------------------------------------------------------------------


Forrester's hierarchy of systems structure (#2 above) really represents a description of the "Structure" level in #1 above.


Combining this with the five step system dynamics modeling process in Chapter 3 of Sterman's (2000) Business Dynamics text, maybe one way to think of a hierarchy of understanding systems (starting from the lowest and moving to highest) might be:


1) ability to translate events into a set of behavior patterns that express a dynamic problem


2) ability to understand an appropriately bounded dynamic hypothesis (either as feedback maps, or stock-flow maps, or a combination, depending on the nature of the problem) for the structure that might be causing the behavior.


3) ability to develop an appropriately bounded dynamic hypothesis (either as feedback maps, or stock-flow maps, or a combination, depending on the nature of the problem) for the structure that might be causing the behavior.


4) ablity to understand a simulation model for testing the dynamic hypothesis.


5) ability to develop a simulation model that can reproduce the dynamic hypothesis.


6) ability to rigorously test the model, both structurally and behaviorally


7) abiliity to use the model to design robust policies.


Notice the distinction between understanding and developing in items 2 and 3, and in items 4 and 5, above.


Of course, much finer levels of understanding of dynamic systems are implicit in the above, e.g.,


- understanding stock-flow dynamics (graphical integration and differentiation)


- understanding basic behaviors (exponential growth and decline, goal-seeking behavior, and oscilllations) and the feedback structures that can cause them,


- understanding how more complex behaviors ( s-shaped growth, s-shaped growth with overshoot and oscillation, and overshoot and collapse), can be produced with combinations of the simple feedback structures,


- understanding of material and information delays and formulations,


- understanding of common decision rule (policy, rate equation) formulations and the importance of bounded rationality in their formulation,


- understanding of common structures, e.g. epidemics and innovation diffusion, path dependence and standards formation, coflows and aging chains, expectations, stock management, and oscillations,


- understanding numerical integration adequately to be able to use it properly,


- understanding the importance of seeking "multiple points of contact between a model and reality by drawing on many sources of data and a wide range of tests" (Sterman's Business Dynamics book, page 850), and


- being able to build models that have "multiple points of contact between...model and reality" and that draw "on many sources of data and a wide range of tests" (ibid),


- etc....


Most of the above was taken from Sterman's (2000) textbook "Business Dynamics: Systems Thinking and Modeling for a Complex World."


Of course, much of the above won't be developed in a five week course, so a much simpler hierarchy may be suitable.


I don't know if any of this helps in identifying a good "hierarchy of qualitatively different ways of understanding dynamic systems," but maybe it will provoke some good conversation....


Paul Newton