Seminar Event Detail

Student AIM Seminar

Date:  Friday, November 10, 2017
Location:  1084 East Hall (4:10 PM to 5:00 PM)

Title:  Differential Privacy and Applications to Game Theory

Abstract:   When social scientists have to ask sensitive or personal questions, there is a real risk that the results will be inaccurate due to dishonest answers. In the 1960's, S. L. Warner came up with a solution. He had participants (privately) flip a coin. If the coin came up heads, they told the truth. If it came up tails, they gave a random answer. He realized that if you gave people plausible deniability then they were more likely to tell the truth.

In the decades since, privacy violations from data analysis have become a much more pervasive problem. The field of privacy-preserving data analytics has developed to help alleviate these concerns, while still supporting meaningful data analysis. In this talk we will be focusing on one particular solution called "differential privacy". Differential privacy has at its core, a similar idea to Weber, that you have maintained someone's privacy if you give them plausible deniability. From a different view, differential privacy gives the guarantee that whatever (positive or negative) consequences that might happen to you as a result of the data analysis are almost equally as likely to occur whether or not your data is included.

We don't only lie because we are embarrassed. Sometimes we lie because the lie will result in someone elses action being more favorable towards us. For example, suppose I am being polled on how much I would pay for an apple. I might report a price below my true price, in the hope that the apple will be sold at the lower price. The pollster would like to design a system that encourages me to report truthfully. Even though I have no desire for privacy in this situation, it turns out that algorithms from the privacy literature can still help design such a system.

In this talk, we'll introduce differential privacy, mechanism design and the connection between them. We'll discuss a mechanism based on differential privacy such that each player is incentivized to tell the truth. This talk is based on this survey paper:


Speaker:  Audra McMillan
Institution:  University of Michigan

Event Organizer:   Nathan Vaughn


Edit this event (login required).
Add new event (login required).
For access requests and instructions, contact

Back to previous page
Back to UM Math seminars/events page.