Preference for Verifiability

Abstract

Decision makers sometimes cannot observe the consequences of their actions ex-post. This paper axiomatically characterizes a decision model in which the decision maker cares about verifying that a good consequence has been achieved. Preferences over acts identify a set of events the decision maker expects to verify. Decision makers choose acts maximizing, in expectation over verifiable events, the worst-case utility consistent with each event. A dual model captures decision makers who instead seek to obscure poor outcomes from verification. As an application, firms choosing carbon-reduction technologies may prefer less efficient but more verifiable technologies to prove emission reductions to stakeholders.