Abstract

Dawid's account of non-empirical theory assessment is meant to complement traditional theory assessments. I contend that his arguments don't provide support for this account. His three arguments, the no alternatives argument, the unexpected explanatory connections argument, and the meta-inductive argument from prior theories' success, are all problematic, particularly for an assessment of string theory. In particular, I argue that the meta-inductive argument is idle, because it's role in underwriting the future predictive success of a theory is subsumed by the normal accounting of its predecessor's predictions in theory growth and testing. Dawid's arguments are interdependent, so showing that one fails is sufficient to cast doubt on his entire account.