Seeing further: The role of modelers and simulation in causal inference

In discussions of causal reasoning, simulation modeling is typically studied as a means of exploring the implications of a set of causal relations. Mathematical modelers manipulate pre-existing causal information, but they do not in general generate new causal discoveries themselves, unless that information was already encoded in the model. This work is reserved for experimenters or observational scientists. However, in certain fields today causal inference, in the sense of discovering novel "off-model" causal relations, is often performed through the aid of computational simulation. Acknowledging this puts more weight on the role of modelers in causal discovery and helps draw attention to the specific properties of simulation which afford discovery processes. In this chapter I identify some of those affordances through a close ethnographic analysis of model-building practices in one computational field, computational systems biology, a field in which modelers typically have little biological expertise and do not generate their own experimental observations.