Center for Social Information Sciences (CSIS) Seminar

Abstract: We investigate the full mechanism design problem for allocating an indivisible object to agents with bi-separable non-expected utility preferences. To model the agent's dynamic preferences, we consider two widely studied frameworks: the recursive utility model and the multi-selves model. Our primary methodological contribution is the introduction of a novel revelation principle for sophisticated non-EU agents. Utilizing this principle, we analyze the canonical revenue maximization problem and establish new connections between the seller's revenue and the allocation rule, extending the classical Myersonian envelope characterization. Furthermore, we demonstrate that (almost) full surplus extraction is achievable under both models. These findings provide valuable insights for extending mechanism design beyond expected utility settings and underscore which instruments social planners should regulate to protect non-expected utility agents from potential exploitation.

Joint work with David S. Ahn.