"The theorist who helped give differential privacy its mathematical teeth"
Aaron Roth is the Class of 1940 Bicentennial Term Associate Professor of Computer and Information Science at the University of Pennsylvania, with a secondary appointment in Statistics and Data Science at the Wharton School. His research centers on the theoretical foundations of machine learning, with an emphasis on privacy, fairness, and algorithmic game theory — including foundational work on differential privacy, the mathematical framework later adopted by the US Census Bureau. Roth has received a Presidential Early Career Award for Scientists and Engineers and multiple best-paper awards for work bridging theoretical computer science and public policy. Alongside longtime collaborator and mentor Michael Kearns, he co-authored The Ethical Algorithm, arguing that many hard problems in algorithmic ethics have precise, provable technical solutions — not just political ones.
Our in-depth summaries and reviews of his work
Kearns and Roth show how to build fairness and privacy directly into algorithms with differential privacy and formal fairness constraints, replacing after-the-fact…
Read Summary →A rigorous mathematical framework for protecting individual data while still learning accurate patterns from it.
Multiple reasonable definitions of algorithmic fairness can be mathematically incompatible with each other.
An algorithm can carry a mathematical proof of its privacy or fairness properties — not just a promise.
"Privacy shouldn't mean choosing between protecting people and learning from data — differential privacy shows you can often do both."— The Ethical Algorithm
"You can't satisfy every reasonable definition of fairness at once — the real question is which tradeoff you're willing to make, and why."— The Ethical Algorithm
Kearns and Roth show how to build fairness and privacy directly into algorithms with differential privacy and formal fairness constraints, replacing after-the-fact policy patches with provable, mathematical guarantees.
