Computer Simulations

The ascent of computer simulations in scientific modelling coincided not only with the rise of the digital computer as such but also with the advent of data visualisation technologies capable of representing the dynamics of a “target system”. In accordance with Mary Hesse’s distinction between formal and material analogies in scientific modelling, computer simulations appear to assume a peculiar dual nature: First, simulations typically are computational realisations of underlying formal models of their target systems, and as such help to determine their empirical correctness. Second, simulations typically comprise an aspect of material modelling, so as to make relevant properties of the target system perceivable. How do these two aspects of computer simulations interact, both on epistemological and pragmatic levels?

Being a relatively recent method of scientific inquiry, computer simulations are an even more recent topic of social and philosophical inquiries into science and technology. There are arguments for their genuinely “motley” epistemological character, for their characteristic “epistemic opacity”, and for their fundamentally dynamic nature, but also for treating them as experiments ‘in silico’. However, these debates only take limited account of simulational practice and its epistemic import. Conversely, from an STS perspective, computer simulations can be understood as complex networks of computer software, hardware, and infrastructures as well as of institutions, communities and practices. However, it remains to be demonstrated how these networks and practices affect the modes of knowledge thus produced. This project aims at an integration of STS and philosophical perspectives in order to address the epistemological and pragmatic aspects of computer simulations as an integrated whole.

The two-part working hypothesis to be tested in the present inquiry is this: First, one model might be realised in a variety of simulational fashions. The formal model bears the primary responsibility for representing the target system, whereas the computational core and the empirical rendering of the simulation are underdetermined by that formal model. Second, the criteria that join the formal and material aspects together are essentially pragmatic. Simulations are chosen in accordance with what shall be communicated about the target system within a community of researchers. In typical contemporary research settings, that community is heterogeneous in disciplinary, topical and technological terms. A computer simulation will be vindicated if and when it is successful on both levels. Underdetermination on the first is resolved on the second level, and problem-oriented, pragmatic empirical adequacy outdoes more foundational epistemological criteria of success.