A Robust Bayesian Truth Serum for Small Populations

Peer prediction mechanisms allow the truthful elicitation of private signals (e.g., experiences, or opinions) in regard to a true world state when this ground truth is unobservable. The original peer method incentive compatible for any number agents n >= 2, but relies on common prior, shared by all and mechanism. Bayesian Truth Serum (BTS) relaxes assumption. While BTS still assumes that share prior need not be known However, only large enough agents, particular required uncertain because it depends prior. In paper, we present robust binary information which every 3, taking advantage particularity quadratic scoring rule. first mechanism provide strict compatibility 3 without relying knowledge Moreover, contrast BTS, our numerically ex post individually rational.