System Dynamics Applications / 2 / 2.7 / 2.7.1

2.7.1 Start Simple—Progressively Add Functionality

Stephen Kline of Stanford University in his 1995 book ‘Foundations of Multidisciplinary Thinking’ comprehensively demonstrates that we humans have cognitive capability seriously outmatched by the complex-world problems we face. Sweeney and Sterman (2000) have demonstrated in ‘Bathtub Dynamics’ that even highly intelligent and well-educated people perform poorly when it comes to predicting dynamic behaviour of the simplest feedback mechanisms. Sterman (2000: 29) explains that people cannot simulate mentally even the simplest feedback system, the first-order linear positive feedback loop. So we should not attempt to build models by assembling the bits with the hope that they will go together and the model will work properly.

Forrester (1994a: 252) stresses:

The starting point for model conceptualisation [as an enabling step to building a quantified system dynamics model] is to first identify the state variables (system stocks, levels or accumulators) then develop the flow rates that cause the state variables to change.

Sterman (2000: 81) goes on to emphasise the importance of quickly building a sufficiently simple working model to foster discussion and initiate critical thinking.

To build models that work as required, it is essential to start small and comprehensively test each module or partial model before adding greater functionality. We must never assume that, when it comes to models of dynamic systems, what simply appears to work, does. It will rarely be so, even for the simplest of models; we must progressively test our models, regardless of how simple they are.

Each module must be treated separately with the interfaces to all other modules and to the outside world clearly defined and documented. Clear and defined steps of systems engineering and software development need to be followed despite the apparent complication and initial set-up delay this brings. Experience, including my own experience in overseeing software development activities, has shown that using this methodology actually speeds up model building by a factor of two or three.

Steps involve:

• defining functional requirements, including interfaces;
• defining objective tests which will demonstrate that the functional requirements will be met by the module and ultimately the completed model;
• validating that this is the right model (or module); and
• verifying that this model (or module) is right, that is, works as designed and sufficiently replicates the reference modes of dynamic behaviour we have specified.

Each module and model contained in this book is explained in detail and applications to support real-world analysis are demonstrated. Examples are coded using Powersim^™ Studio. More advanced modules and models are also included along with the explanatory scenarios, to which they apply. As far as possible the example modules and models have been selected for their real-world applicability.

A trial version of Powersim^TM Studio can be downloaded from the Powersim^TM website, http://www.powersim.com.