14h00 - 15h00
The design of incentive mechanisms through statistical machine learning
Mechanism design studies situations where a set of agents each hold private information regarding their preferences over different outcomes. A mechanism receives claims about preferences and chooses an outcome and payments. In this talk, I draw a connection between the discriminant functions of multi-class classifiers and incentive-compatible mechanism design, enabling new applications of optimization for the design of mechanisms in settings both with and without money, including auctions, assignment and matching.