Fischer on backtracking and Newcomb's problem

Abstract

The right solution to Newcomb’s problem of often thought to depend on the plausibility of counterfactual backtracking. If the predictor’s prediction is counterfactually dependent on my choice, it seems rational to choose only the second box. If there is no such counterfactual dependence, the dominance argument appears to vindicate the two-box solution. John Martin Fischer has challenged this wisdom. He claims that the two-boxer can grant the truth of the relevant backtracking counterfactuals, while sticking to the dominance argument for choosing both boxes. In this note I try to show that his argument, although ingenious, does not succeed.