Lies and consequences

I study a strategic-communication game between an informed sender and an uninformed receiver with partially aligned preferences. The receiver is endowed with the ability to probabilistically detect if the sender is lying. Specifically, if the sender is making a false claim about her type, with some commonly known probability p the receiver additionally observes a private signal indicating that the sender is lying. The main result is that the receiver’s stochastic lie-detection ability makes fully revealing equilibria—the best outcome for the receiver—possible, even for small p (less than $$\frac{1}{2}$$ 1 2 ). Additionally, if the language consists of precise messages, fully revealing equilibria exist only for $$p=1$$ p = 1 and for a set of intermediate values of p that is bounded away from 0 and 1, making the maximal ex-ante expected equilibrium utility of the receiver non-monotone in p. If vague messages are allowed, full revelation can be supported for all large enough p, overturning the non-monotonicity and improving communication outcomes relative to the precise-language case.