Predictive Infelicities and the Instability of Predictive Optimality
Abstract
Recent neo-Humean theories of laws of nature have placed substantial emphasis on the characteristic epistemic roles
played by laws in scientific practice. In particular, these theories seek to understand laws in terms of their optimal
predictive utility to creatures in our epistemic situation. In contrast to other approaches, this view has the distinct
advantage that it is able to account for a number of pervasive features possessed by putative actual laws of nature.
However, it also faces some unique challenges. First, since the view tries to characterize the laws in terms of their
predictive utility, any respects in which putative actual laws are sub-optimally predictively useful are inherently
problematic. Such "predictive infelicities" can easily be found among our best physical theories. Second, in tying the laws
to our epistemic situation, neo-Humeanism raises the possibility that by changing our epistemic situation, we can change the laws themselves, though it is hard to believe that we have such influence over the laws. This paper aims to address both of these challenges, first by presenting a variety of strategies for explaining away predictive infelicities, and then by developing a version of Lewis's "rigidification" strategy to preclude the possibility of our changing the laws.