Computational templates as representations

Paul Humphreys, Philosophy, University of Virginia

For some types of complex systems a specific form of mathematical representation (e.g. network theory, fitness landscapes, Ising models) can be used to represent very different subject matters. Computationally, this is a huge advantage but it poses well-known problems for realistic interpretations. It also suggests that theories, models, and simulations of these systems stand in different relationships than do theories and models in more traditional areas. I shall draw on some ideas from my 2004 book on computational science (Extending Ourselves) and relate them to some recent collaborative modeling work exploring constrained searches on fitness landscapes. Some tentative conclusions will be drawn about the status of computational templates.