Equilibrium Selection in Information Elicitation without Verification via Information Monotonicity

In this paper, we propose a new mechanism – the Disagreement Mechanism – which elicits privately-held, non-variable information from self-interested agents in the single question (peerprediction) setting. To the best of our knowledge, our Disagreement Mechanism is the first strictly truthful mechanism in the single-question setting that is simultaneously: Detail-Free: does not need to know the common prior; Focal: truth-telling pays strictly higher than any other symmetric equilibria excluding some unnatural permutation equilibria; Small group: the properties of the mechanism hold even for a small number of agents, even in binary signal setting. Our mechanism only asks each agent her signal as well as a forecast of the other agents’ signals. Additionally, we show that the focal result is both tight and robust, and we extend it to the case of asymmetric equilibria when the number of agents is sufficiently large. 1998 ACM Subject Classification J.4 Social and Behavioral Sciences