Image courtesy of Invited Speakers
Constantinos Daskalakis (MIT)
Title: Mechanism Design for Learning Agents
Abstract: Combinatorial auctions, the task of allocating items to strategic buyers with combinatorial valuations over bundles, has been a paradigmatic problem for mechanism design. The celebrated VCG mechanism solves the problem, but it is expensive in computation and communication, motivating the search for alternative mechanisms. We overview the advances of algorithmic mechanism design on this subject, and propose an online-learning approach to mechanism design, sidestepping impossibility barriers that have been identified for buyers with submodular valuations. (Based on joint work with Vassilis Syrgkanis)
Olivier Gossner (LSE and École Polytechnique)
Title: A Normalized Value for Information Purchases
Abstract: Consider agents who are heterogeneous in their preferences and wealth levels. These agents may acquire information prior to choosing an investment that has a property of no-arbitrage, and each piece of information bears a corresponding cost. We associate a numeric index to each information purchase (information-cost pair). This index describes the normalized value of the information purchase: it is the risk-aversion level of the unique CARA (Constant Absolute Risk Aversion) agent who is indifferent between accepting and rejecting the purchase, and it is characterized by a “duality” principle that states that agents with a stronger preference for information should engage more often in information purchases. No agent more risk-averse than the index finds it profitable to acquire the information, whereas all agents less risk-averse than the index do. Given an empirically measured range of degrees of risk aversion in an economy with no-arbitrage investments, our model therefore comes close to describing an inverse demand for information, by predicting what pieces of information are acquired by agents and which ones are not. Among several desirable properties, the normalized-value formula induces a complete ranking of information structures that extends Blackwell’s classic ordering. (Based on joint work with Antonio Cabrales and Roberto Serrano)
Kurt Mehlhorn (Max-Planck-Institut für Informatik)
Title: Linear Fisher Markets with Satiation
Abstract: We discuss linear Fisher markets with satiation. In these markets, either buyers have utility caps (upper bounds on the amount of utility they want to achieve) or sellers have earning caps (upper bounds on the amount of money they want to earn), or both buyers and sellers have caps. In the former two cases, equilibria can be characterized by convex programs and polynomial time algorithms exist by virtue of the Ellipsoid method. Polynomial-time combinatorial algorithms also exist. In the latter case (utility and earning caps), the set of equilibria can be non-convex. The talk is based on three recent papers by Nikhil Devanur, Kamal Jain, Tung Mai, Vijay Vazirani, and Sadra Yazdanbod; Xiaohui Bei, Jugal Garg, Martin Hoefer, and myself; and Jugal Garg, Martin Hoefer, and myself.
