The strategy of model-based science

My title refers to Richard Levins’ famous paper on models in population biology (1966). Here Levins presented his three-way distinction between kinds of modelbuilding, and also introduced a set of more fundamental ideas about trade-offs that constrain and guide scientific work. For Levins, these trade-offs derive from the relationships between three different theoretical goals: realism, precision, and generality. The talk of ‘‘strategies’’ within Levins’ paper concerns alternative strategies within the enterprise of model-building. My topic here is broader; I will treat models and model-building as characteristic of one particular approach to theorizing, a strategy of model-based science. The ideas presented here are indebted to Michael Weisberg’s work (forthcoming, 2003), and to discussions with him at Stanford. They are also indebted to a scattered tradition of other philosophical and scientific work, especially by Levins, Giere, and Wimsatt. The aim of the paper is to give a general sketch of model-based science as I see it, but a sketch focusing particularly on divergences between myself and those on whose work I draw. The term ‘‘model’’ is surely one of the most contested in all of philosophy of science. What sense of the term do I have in mind? My starting point is the account given by Giere (1988). For Giere, models are idealized structures that we use to