TechFest Workshop – Theory Day – Session 1

• March 10, 2016
• Bobby Kleinberg | Microsoft

We explore the misalignment of incentives in systems that use crowdsourced effort to solve large problems made up of smaller tasks, for example aggregating millions of reviews to create a system for recommending products or news articles. In most such systems one cannot directly assign tasks to workers, but can only observe a crowd going about their everyday activities. The system designer’s motivation (to efficiently explore a space of alternatives) is thus in conflict with that of the crowd’s individual members (to derive utility from the alternatives they personally select). We model this as a type of multi-armed bandit problem and present a full characterization of the trade-off between the principal’s expected payments and the crowd’s total reward. This is joint work with Peter Frazier, David Kempe, and Jon Kleinberg.