Bowen He

Standard accounts of higher-order defeat run into well-known Bayesian hurdles. In this paper, I explore the prospects of overcoming those hurdles by applying a thoroughgoing instrumentalist approach: treating our reasoning capacities as we’d treat any other instrument. I argue that instrumentalism cannot coherently accommodate the defeatist’s position when that position involves genuine higher-order uncertainty—that is, rational uncertainty about what rationality requires. At the end of the paper I gesture at an approach to modeling our responses to defeaters that is both instrumentalist and Bayesian, but that abandons the thought that the cases typically described as cases of higher-order defeat involve genuinely higher-order uncertainty.

A popular version of epistemic Permissivism says that, given the total evidence, sometimes there is a permissible credence range towards a proposition. Ginger Schultheis (2018) offers a Dominance Argument against it. Schultheis argues that it is irrational to hold a credence at the edge of any permissible range because the edge credence takes higher risks of being irrational than the credence in the middle. In this paper, I propose two new responses. Firstly, I argue that after the risk assessment in irrationality, a new stable range may emerge such that each credence from it does not take more risks than others. Schultheis’s Dominance Argument can only shrink the original credence range to this new stable range. Second, I argue that sometimes it is rational for us to hold a more risky credence when a safer alternative is available. If rationality aims at truth-conduciveness and informativeness, a credence’s higher risks of being irrational do not render it irrational when one risks being less truth-conducive in exchange for informativeness.