Comments on “On the very idea of a system dynamics model of Kuhnian science”

Abstract

Wittenberg notes some differences between models of theories and models of real systems and then concludes that validation of these two types of models must be fundamentally different: Behavior validation, which is important in validating models of real systems, loses its meaning in validating models of theories, where emphasis must shift structural validity. My first disagreement with this argument is its implication that for models of real systems emphasis is normally on behavior validity. In fact, since system dynamics models are causal ones, structural validity must always be crucial, whether one deals with models of theories or models of real systems. Without structural validity, behavior validity is meaningless for any system dynamics model, by definition. (This is not true for noncausal, statistical models.) I therefore find superfluous, even misleading, Wittenberg's suggestion that emphasis be shifted evaluating the structural validity of models of theories. My other objection Wittenberg's conclusion is its suggestion that behavior validity is spurious for all models of theories, by their nature. This is a false generalization. It is often possible (sometimes necessary) carry out tests of behavior validity on models of theories. In fact, the boundary adequacy test that Wittenberg himself performs on Sterman's model is an important counterexample. Since a major stated purpose of Sterman's model is test the dynamic consistency of Kuhn's theory, behavior validity becomes crucial! If we argue as Wittenberg suggests, that behavior validity of Sterman's model is spurious by definition (because we have no objective access the alleged behavior), then this means that the model makes a major claim tbat is impossible test, which is tantamount declaring tbe model void. The only way evaluate the dynamic consistency of any theory is compare its alleged behavior with its deduced behavior. Since Wittenberg states that the purpose of models of theories is in general to probe an argument's internal consistency, bis conclusions about behavior validity would render all models of theories void, in principle. My disagreement with Wittenberg's distinction between validating models of real systems and validating models of theories should not imply that I am against any such distinction. One could come up with a number of practical differences