Abstract

In this paper, I provide an accuracy-based argument for conditionalization (via reflection) that does not rely on norms of maximizing expected accuracy. (This is a draft of a paper that I wrote in 2013. It stalled for no very good reason. I still believe the content is right).