Imagine auditing a bank for potential discrimination. If White loan recipients were found to repay their loans less often than racial minorities, it would suggest a double standard, with a relatively lenient bar applied to White applicants. Such an “outcome test” is one of the most popular empirical strategies for detecting discrimination. Outcome tests are simple to apply and their logic seemingly simple to understand. But almost immediately after their introduction 30 years ago, outcome tests were found to suffer from deep theoretical limitations. Researchers have since worked to reconcile the intuitive appeal of outcome tests with their formal limitations, though with limited success. In this talk, I'll explain how a small tweak to standard outcome tests yields surprisingly strong statistical guarantees while preserving the logistical and rhetorical simplicity of the traditional approach. Applying our “robust outcome test” to 2.8 million police stops across California, we find evidence of pervasive discrimination — a pattern that would have been missed by the standard outcome test.
Sharad Goel is a Professor of Public Policy at Harvard Kennedy School. He looks at public policy through the lens of computer science, bringing a computational perspective to a diverse range of contemporary social and political issues, including criminal justice reform, democratic governance, and the equitable design of algorithms. Sharad is the founder and co-director of the Computational Policy Lab, an interdisciplinary team of researchers, data scientists, and journalists that use technology to drive social impact. Prior to joining Harvard, Sharad was on the faculty at Stanford University, with appointments in management science & engineering, computer science, sociology, and the law school. Sharad holds an undergraduate degree in mathematics from the University of Chicago, as well as a master’s degree in computer science and a doctorate in applied mathematics from Cornell University.
LOCATION:SEC 3.301/302/303 STATUS:CONFIRMED DTSTART:20250224T163000Z DTEND:20250224T173000Z END:VEVENT END:VCALENDAR