Michael Albert is Assistant Professor of Business Administration in Darden's MBA program and holds joint appointments in Systems Engineering and Computer Science in the School of Engineering and Applied Sciences. The title of his talk is Machine Learning and Market Design.
Many online marketplaces (such as auctions for online advertising) are continuously optimized in response to the behavior of the market participants. This continuous optimization can be viewed as a machine learning algorithm that takes as inputs the actions of the market participants and outputs a particular structure for the market. In this talk, I will discuss the role of algorithms in market (or mechanism) design and the successful applications of these algorithms to modern markets. I will also present some work that I have done on mechanism design with learning in a setting where learning provides significant advantages. Finally, I will talk about how strategic considerations of the market participants can potentially cause this learning process to backfire and propose a potential path forward to learn in a strategy-proof manner.