Juba Ziani is a Warren Center Postdoctoral Fellow at the University of Pennsylvania, hosted by Sampath Kannan, Michael Kearns, Aaron Roth, and Rakesh Vohra. Prior to this, he was a PhD student at Caltech in the Computing and Mathematical Sciences department, where he was advised by Katrina Ligett and Adam Wierman. Juba studies the optimization, game theoretic, economic, ethical, and societal challenges that arise from transactions and interactions involving data. In particular, his research focuses on the design of markets for data, on data privacy with a focus on "differential privacy", on fairness in machine learning and decision- making, and on strategic considerations in machine learning.
Data is now everywhere: enormous amounts of data are produced and processed every day. Data is gathered,exchanged, and used extensively in computations that serve many purposes: e.g., computing statistics on populations, refining bidding strategies in ad auctions, improving recommendation systems, and making loan or hiring decisions. Yet, data is not always transacted and processed in a responsible manner. Data collection often happens without the data holders' consent, who may also not be compensated for their data. Privacy leaks are numerous, exhibiting a need for better privacy protections on personal and sensitive data. Data-driven machine learning and decision making algorithms have been shown to both mimic past bias and to introduce additional bias in their predictions, leading to inequalities and discrimination. In this talk, I will focus on my research on using data in a more responsible manner. The main focus of the talk will be on my work on the privacy issues that arise in data transactions and data-driven analysis, under the lens of a framework known as differential privacy. I will go over my work on designing transactions for data where we provide differential privacy guarantees to the individuals whose sensitive data we are buying and using in computations, and will focus on my recent work on providing differential privacy to agents in auction settings, where it is natural to want to protect the valuations and bids of said agents. I will also give a brief overview of the other directions that I have followed in my research, both on the optimization and economic challenges that arise when letting agents opt in and out of data sharing and compensating them sufficiently for their data contributions, and on how to reduce the disparate and discriminatory impact of data-driven decision-making.
--
Please contact MPI-SWS office team for Zoom link information