Auctions between Regret-Minimizing Agents: on Learning Dynamics and User Incentives (via Zoom)

Abstract: The usage of automated learning agents is becoming increasingly prevalent in many online economic applications such as online auctions and automated trading. In this talk we will discuss the strategic situations that the users of such automated learning agents are facing. The talk is based on two joint works with Noam Nisan: Auctions between Regret-Minimizing Agents (WWW 2022) and How and Why to Manipulate Your Own Agent: on the Incentives of Users of Learning Agents (NeurIPS 2022). In these works, we propose to view the outcomes of the agents’ dynamics as inducing a “meta-game” between the users. Our main focus is on whether users can benefit in this meta-game from “manipulating” their own agents by misreporting their parameters to them.  We analyze a scenario in which software agents implemented as regret-minimizing algorithms engage in a repeated auction on behalf of their users. We study first-price and second-price auctions, as well as their generalized versions (e.g., as those used for ad auctions). Using both theoretical analysis and simulations, we show that, surprisingly, in second-price auctions the players have incentives to misreport their true valuations to their own learning agents, while in first-price auctions it is a dominant strategy for all players to truthfully report their valuations to their agents.  
 

Bio: Yoav Kolumbus has recently joined the Center for Data Science for Enterprise and Society at Cornell as a research assistant professor, with the mentorship of Prof. √Čva Tardos, Prof. Robert Kleinberg, and Prof. David Easley. Prior to that, Yoav completed his Ph.D. in Computer Science at the Hebrew University of Jerusalem, advised by Prof. Noam Nisan, and his M.Sc. in Physics, advised by Prof. Sorin Solomon. His research interests lie at the interface of machine learning, algorithmic game theory, dynamical systems, and networks, with a focus on interactions between learning and strategic play in multi-agent systems.