Inbal Talgam Cohen
The standard approach to maximizing revenue in economic mechanism design is Bayesian. This approach assumes that buyer preferences are drawn from prior distributions and that these distributions are known to the mechanism. We design alternative mechanisms that extract approximately optimal revenue without knowledge of the priors. This serves two goals: First, the resulting mechanisms are robust to changes in the preference distributions, and relax idealized common-knowledge assumptions (i.e., adhere to the Wilson doctrine). Second, the settings we consider are complicated – revenue maximization with multiple kinds of goods or with uncertainty of buyers regarding their own preferences (interdependent values) – and the prior-independence approach results in simple designs that contribute to our understanding of these settings.
Joint work with Tim Roughgarden and Qiqi Yan