UC Irvine

Contemporary astrophysics and cosmology, like other areas of science that study remote or difficult to manipulate subjects, must often rely on sophisticated computer simulations. These simulations can enable investigations that would otherwise be impossible, but they also come with costs: the numerical methods used raise the spectre of numerical errors, which in turn threatens the reliability of the simulations themselves. And while some sources of numerical error are well-understood and can, in some circumstances, be accounted for, the highly nonlinear nature of these simulations makes it exceedingly difficult to categorically rule out all possible sources of error. In this dissertation, I develop and defend an epistemic framework for thinking about these simulations which is informed by scientific practice in astrophysics and cosmology and corrects deficiencies in previous philosophical accounts.