Abstract

Suppose we consider an agent with both numerical credences and all-or-nothing beliefs. This agent might also have a plan about how she is going to update her beliefs upon receiving new evidence. What rational requirements on such a plan can be justified from an epistemic value point of view? Plan Almost Lockean Revision is the claim that it is rationally required that one’s planned beliefs are exactly one’s sufficiently high conditional credences. We start by reviewing arguments available for Plan Lockean Revision in the current literature, ultimately concluding that they are non-optimal. We provide a better argument to the effect that the belief updating rule that is expected to be the best according to one’s current credences is exactly Plan Almost Lockean Revision, that is, we prove a Qualitative Greaves-Wallace Theorem. Furthermore, building on the work of (Rothschild, 2021), we investigate the dutchbookability of Lockean betting behavior for all-or-nothing beliefs and their plannings, ultimately proving a qualitative version of the dutch strategy theorem which leads to the development of novel dutch-strategy/accuracy-dominance arguments for Lockean norms on belief/belief-planning pairs.