LAUSR

The details vary, but the central idea is that consequence for epistemic language should track instead a notion of preservation of support by an information state. On the resulting view, a conclusion B follows from a set of premises A1, . . . ,An just in case all bodies of information that support A1, . . . ,An also support B. This view of consequence naturally dovetails with a non-truthconditional semantics, on which epistemic modal claims don’t express propositions and are not true or false. This paper investigates the link between informational consequence and credence. I first suggest a natural and seemingly harmless constraint concerning this link: informational consequence is certainty preserving. I.e., on any rational credence distribution, when the premises of an informational inferences have credence 1, the conclusion also has credence 1. This constraint has never been explicitly defended (though Kolodny and MacFarlane 2010 come close to doing so), but it dovetails with the widespread view that informational consequence tracks preservation of acceptance. It also allows us to make sense of intuitive judgments about informational inferences, including McGee-style alleged counterexamples to Modus Ponens. After this setup, I show that, unfortunately, the certainty-preserving constraint leads to triviality. In particular, we can show that the following three claims are incompatible: (i) informational consequence is extensionally distinct from classical consequence (in particular, some inferences that are informationally valid are classically invalid); (ii) informational inferences are