Abstract
In the first part of the talk, I’ll describe work with Chawla, Devanur and Sivan on a pricing problem where a buyer is interested in purchasing/using a good, such as an app or music or software, repeatedly over time. A common assumption in auction theory and mechanism design is that a consumer knows how much he values an item at the moment he is considering a purchase and that he has an accurate prior over how much he will value that item in the future (if not precise knowledge). In this work, we explore scenarios where the consumer discovers his valuation for a good as he uses it, and his value per usage evolves as a martingale. We show how to construct simple pricing mechanisms that yield approximately optimal seller revenue regardless of the buyer’s risk profile. In the second part of the talk, I’ll describe results with Fiat, Goldner and Koutsoupias on how to maximize auctioneer revenue when he is selling a service for which buyers have both a value and a deadline. In this setting, we show how to use linear programming duality to design the optimal auction.