Full Program »
Simplicity and Unification in Cosmological Model SelectionThe criteria for model selection commonly used in cosmology tend to focus on optimizing the number of free parameters, so as to avoid overfitting. This is often justified by appealing to a sense of simplicity or ontological parsimony. I claim that, strictly speaking, it is not simplicity that these model selection criteria are after. Moreover, I show that statistical model selection methods in general are particularly problematic in cosmology due to the limited theoretical background and possibility for experimentation. I explore whether a Bayesian criterion of unification can be better suited to some probabilistic inferences in cosmology.
Center for Philosophy of Science
University of Pittsburgh