Seminar on History and Philosophy of Science

Abstract:  How should an agent act in the face of normative uncertainty? A natural thought: we should generalize a popular theory for choices under merely descriptive uncertainty: expected utility theory. But this decision rule faces a serious challenge in the context of normative uncertainty: it requires the utilities assigned by different moral theories to be comparable. This paper defends a strategy for avoiding problems associated with intertheoretic utility comparison, using utilities specified de dicto. I provide a decision theory for de dicto utilities that generalizes expected utility theory but does not require intertheoretic utility comparisons. I show how the strategy generalizes to what I call "the hard problem of intertheoretic utility comparisons": the problem of normative uncertainty about whether the correct moral theory has cardinal or merely ordinal structure. I argue that these theories should be treated as falling on a spectrum; intermediate theories on this spectrum carry only partial cardinal information. This motivates representing both merely ordinal and intermediate theories with imprecise utilities. I show that familiar decision theories for imprecise credences face a problem in the context of normative uncertainty, and introduce a class of novel decision rules that avoid the problem.