Recent literature on causal decision theory (CDT) has featured much discussion of what Hare & Hedden call ``decision dependence''---the fact that, for a causalist, the expected utility of an act a can sometimes depend on how confident one is that one will perform a. This is widely seen as puzzling. In this talk, I will focus on decision dependent cases in which CDTers believe that they are subject to tragic evidential correlations (henceforth TECs). In a TEC case, the more confident a CDTer grows that she will perform a given act a, the more confident she becomes that a is a bad option: there is a tragic correlation between her confidence that something is her next move and the expected utility she assigns to it.
Unlike the proponent of evidential decision theory (EDT), the CDTer distinguishes between tragic correlations that are merely evidential, and those that are causal as well. This leads the CDTer to have a complex attitude towards her own anticipated pessimism in TEC cases. As James Joyce puts it, in a TEC case, the CDTer "[does] not fully trust the accuracy of the future beliefs on which [her] regrets about [acting] will be based''. I present a decision problem in which this extra bit of doubt, on behalf of the CDTer, results in her making better decisions than the EDTer.