QUALITATIVE SIMULATION AND SYSTEM DYNAMICS

An objective of this work is to move toward integrating qualitative simulation (QS) procedures with system dynamics models. As a complement to the studies of qualitative analysis that are at present being performed from the mathematical point of view, it would be interesting to establish a way of incorporating QS techniques into system dynamics. Because of the graphical representation and causal definition of its models, system dynamics is included among the structural modeling techniques (Lendaris 1980). QS is a special kind of simulation for qualitative reasoning about physical systems (Kuipers 1984; 1986; 1987). System dynamics and QS have the underlying concepts of structural modeling in common: a structural model expressed as a diagram consisting of a set of nodes (representing variables) and their interconnections. Both assume that a model can be defined in a graphical manner. Emphasis in both is on the qualitative structural aspects of a system to be modeled rather than on its numerical or statistical properties. Both therefore aim at representing the system by means of a diagram and later simulating the system numerically (system dynamics) or qualitatively (QS). First, the propagation of constraints (the basis of QS algorithms) is studied from the point of view of constraint theory (Friedman 1967; 1976; Friedman and Leondes 1969). I show why the structure of the Forrester diagrams can establish a general scheme that would be valid for the propagation of constraints. The modeling proposal is based on these results. By maintaining this structure throughout the modeling process, all structural errors can be avoided. Second, I examine some similarities and differences between the system dynamics and QS methodologies, and select those characteristics that permit integration. The software designed for such integration is also explained. Finally, I examine a macroeconomic model and comment on some of the results observed in the QS literature.