Time | Talk |
---|---|

9:00 - 10:00 | Jason Hartline Introduction to Bayesian Mechanism Design [slides] |

10:00 - 10:30 | Tim Roughgarden Simple/Prior-Independent Auctions (Part A) [slides] |

Time | Talk |
---|---|

11:00 - 11:30 | Tim Roughgarden Simple/Prior-Independent Auctions (Part B) |

11:30 - 12:30 | Costis Daskalakis Revenue-optimization in multi-item auctions [slides] |

Time | Talk |
---|---|

2:30 - 3:30 | Nicole Immorlica Black-box reductions from mechanism to algorithm design [slides] |

3:30 - 4:00 | Shuchi Chawla Non-linear objectives in mechanism design (Part A) [slides] |

Time | Talk |
---|---|

4:30 - 5:00 | Suchi Chawla Non-linear objectives in mechanism design (Part B) |

5:00 - 6:00 | Eva Tardos Price of Anarchy of Practical Auctions [slides] |

Simple/Prior-Independent Auctions

Speaker: Tim Roughgarden (Stanford)

Optimal mechanisms are often complex and depend on the details of the distribution of bidders' preferences. On the other hand, strong impossibility results hold under worst-case input assumptions, as discussed above. To mediate between these extremes the literature has proposed interesting models bridging worst- and average-case models, and compelling positive results have been obtained for such models. Examples include simple, approximately optimal mechanisms, which are only parameterized by minimal properties of the distribution; and prior-independent auctions, whose guarantees hold under the assumption that the agents' preferences are stochastic, but without knowledge of the underlying distributions of preferences. This lecture will survey work on this front.

Revenue-optimization in multi-item auctions

Speaker: Costis Daskalakis (MIT)

While welfare optimizing mechanism design has been solved in very general settings, even under worst-case assumptions about the agents' preferences, research on revenue optimization has been stagnant. Myerson's seminal work provides a revenue-optimizing single-item auction, but generalizing this auction to more general settings, e.g. multi-item auctions, has been challenging to economists. Algorithmic techniques have enabled breakthroughs on this problem, enabling exactly optimal mechanisms. This lecture will survey work on designing revenue optimal auctions, discuss the relevant combinatorial optimization techniques, and present open problems going forward.

Black-box reductions from mechanism to algorithm design

Speaker: Nicole Immorlica

Computational considerations aside, welfare optimizing mechanism design has been solved, even under worst-case assumptions about the bidders' preferences. It has also been shown that, under worst-case assumptions, not all welfare problems that can be efficiently approximated can also be approximated in a mechanism design setting. On the other hand, Bayesian information about the agents' preferences enables very general, black-box reductions from mechanism to algorithm design. This lecture will survey work on this frontier, outlining challenges in going beyond the fully Bayesian model and the welfare objectives.

Non-linear objectives in mechanism design

Speaker: Shuchi Chawla (University of Wisconsin-Madison)

The mechanism design literature commonly focuses on the welfare and revenue objectives. Less is known for other objectives and typically results are negative, e.g. for minimizing makespan in scheduling jobs to strategic machines. Here, too, the Bayesian perspective has provided ways to circumvent worst-case impossibility results. This lecture will survey work on mechanism design with general objectives.

Price of Anarchy of Practical Auctions

Speaker: Eva Tardos (Cornell)

Auctions used in practice, such as the generalized second price auction of sponsored search, are often not optimal theoretically. Recent work has quantified the loss in objective, e.g. welfare, of using such suboptimal auctions. This lecture will survey this work, outlining challenges in analyzing incomplete information settings.