Systems Science Seminar

An information theoretical view of information elicitation mechanisms

Yuqing Kong

University of Michigan
Thursday, June 08, 2017
4:00pm - 5:00pm
1005 EECS

Add to Google Calendar

About the Event

An important question in crowdsourcing is how to elicit accurate and useful information without verification. Starting from Miller, Resnick, and Zeckhauser [2004], peer prediction style mechanisms become popular and important information elicitation tools when the information cannot be (cheaply) verified. This talk will employ information theory to provide an information theoretical mechanism design framework for peer prediction. This talk will also recast several important prior results (Miller, Resnick, and Zeckhauser [2004], Bayesian Truth Serum [2004], Dasgupta and Ghosh [2013]) into this information theory framework and apply the framework to design several new peer prediction mechanisms. This new approach promises to provide a systematic theoretical understanding for peer prediction.

Additional Information

Contact: Judi Jones

Phone: 763-8557


Sponsor(s): ECE - Systems

Faculty Sponsor: Achilleas Anastasopoulos

Open to: Public