Multi-Dimensional Screening: Buyer-Optimal Learning and Informational Robustness

What is the optimal mechanism that a monopolist should use to sell multiple goods to a single buyer? Despite being a classic economic problem, multi-dimensional screening is notoriously intractable. Even if the seller has just two goods and the buyer's values are additive, independent, and identically distributed, the optimal mechanism is hard to characterize generally. In this paper, we study a general version (with arbitrarily many goods and non-additive values) of this problem but with the novel feature of buyer learning. As it turns out, introducing this new feature makes the model tractable and in certain environments---including the one with independent and additive values---makes pure bundling an optimal mechanism.