Attempting to detect a lie: do we think it through?

Abstract Game-theoretic analyses of communication rely on beliefs—especially, the receiver’s belief about the truth status of an utterance and the sender’s belief about the reaction to the utterance—but research that provides measurements of such beliefs is still in its infancy. Our experiment examines the use of second-order beliefs, measuring belief hierarchies regarding a message that may be a lie. In a two-player communication game between a sender and a receiver, the sender knows the state of the world and has a transparent incentive to deceive the receiver. The receiver chooses a binary reaction. For a wide set of non-equilibrium beliefs, the reaction and the receiver’s second-order belief should dissonate: she should follow the sender’s statement if and only if she believes that the sender believes that she does not follow the statement. The opposite is true empirically, constituting a new pattern of inconsistency between actions and beliefs.