(Draft Coming Soon!) I study a prior-free game in which agents can learn coarse information about a type distribution before participating in a multidimensional scoring auction. I develop a notion of strategic simplicity, asking when this coarse information is sufficient for agents to play equilibrium strategies. I find that satisfaction of this criterion relates to the structure of the auction's underlying equilibrium. When this structure is sufficiently well-behaved, my simplicity notion is satisfied and other nice properties of the auction are guaranteed. I exploit one such property, a payoff equivalence result, to develop a simple condition to test whether a scoring auction admits an equilibrium with coarse priors.

(with Ben Brooks and Songzi Du) We characterize robustly optimal mechanisms in single-unit auctions in which the seller knows the mean and support of each buyers' value distribution but does not know the correlation between players' values or their signal structures. We identify worst-case information structures and find general conditions on allocation rules that guarantee optimality. We find that in the symmetric case, a proportional auction satisfies these conditions. In the asymmetric case, we develop a generalization of the proportional auction and show that it retains some of the latter's appealing properties.

(with Nicole Immorlica, Brendan Lucier, and Clayton Thomas)