The paper investigates what is meant by "good science" and "bad science" and how these differ as between the natural (physical and biological) sciences on the one hand and social sciences on the other. We conclude on the basis of historical evidence that the natural science are much more heavily constrained by evidence and observation than by theory while the social sciences are constrained by prior theory and hardly at all by direct evidence. Current examples of the latter proposition are (...) taken from recent issues of leading social science journals. We argue that agent based social simulations can be used as a tool to constrain the development of a new social science by direct (what economists dismiss as anecdotal) evidence and that to do so would make social science relevant to the understanding and influencing of social processes. We argue that such a development is both possible and desirable. We do not argue that it is likely. (shrink)

The classical view of "rational man" as the unit of analysis for economic behavior and marketplace exchange has been changed by the late twentieth century with the help of behavioral economics that considers predictable irrationality as a normal mode of behavior. Instead of revising neoclassical economics to fit contemporary economic crises, it is recommended to follow Adam Smith's original concerns for the social setting of individual behavior and to treat economic crises with pragmatic flexibility rather than with dogmatic ideology.