Theory Seminar

Peer Prediction and Mechanism Design

Yuqing KongPhD StudentUniversity of Michigan

When we collect self-report surveys from people, people may not tell the truth. However, if we assume each person's private information is "informative" about his peers' private information, we can always use a framework called "peer prediction" to motivate people to tell the truth through additional payments, even when we cannot directly verify an agents answer. This talk will give a survey of related results using peer prediction based mechanisms in different settings.
For example, if agents are asked to rate a restaurant, settings might include: when agents share an initial expectation about the restaurant, when the mechanism know this initial expectation as well, and when there are many or few agents.

Sponsored by