AI Seminar

Practical Issues in Mechanism Design

Yevgeniy "eugene" Vorobeychik

Mechanism Design has enjoyed considerable success, both theoretically (by providing general insights) and practically. However, even the ultimate practical successes have been plagued by difficulties of reconciling the theoretical ideals with practical limitations. Some of these limitations have to do with computational issues, such as complexity of preference elicitation. Others are side-effects of practical constraints. In designing real mechanisms, theory would thus provide a sensible starting point at best, and often merely general intuition. Thereafter, ad hoc experiments and back-of-the-envelope calculations would ensue. Whether these in the end result in an optimal mechanism is usually unclear, and often serious flaws remain through several design iterations. In this talk, I attempt to outline some of the common difficulties that arise when theory meets practice, and suggest an automated (approximate) mechanism design framework that may go a long way to remedy these.

Sponsored by

AI Lab