Automated Market Design via Reinforcement Learning

Date: 

Friday, April 1, 2022, 1:00pm to 2:30pm

Location: 

SEC 1.413 & on Zoom here: https://harvard.zoom.us/j/97692964231?pwd=K05BMEhDNTZtbUhHYkZ5S21qZ2FKQT09#success

Speaker:  Gianluca Brero

Title:  Automated Market Design via Reinforcement Learning

Abstract:  In this talk, I will introduce the use of reinforcement learning to design algorithmic markets. First, I will show how (deep) reinforcement learning algorithms can be used to design sequential price mechanisms, where we can assume truthful agents’ behavior. I will then consider the more general class of mechanisms that combine a messaging round with a sequential-pricing stage. The rules of the sequential-pricing stage—and in particular the way these rules use messages—determine the way the messaging stage is used. This is a Stackelberg game where the designer is the leader and fixes the mechanism rules, inducing an equilibrium amongst agents (the followers). I will introduce a novel single-agent Stackelberg MDP formulation, where the leader learns to effect a follower equilibrium that optimizes its objective. In the last part of my talk, I will show how the Stackelberg MDP framework can be used to design economic platforms that prevent collusive behaviors of AI pricing algorithms.