The Problem of Prediction - Social Justice for Data Science

The Problem of Prediction Date: Tuesday, October 21, 3:30-5:00 PM Location: BU Computing & Data Sciences Room 1101, 665 Commonwealth Avenue, Boston, MA (virtual) Abstract: Prediction is the lifeblood of artificial intelligence (AI). When a bank denies a mortgage, it relies on AI to statistically predict creditworthiness and the likelihood of on-time repayment. When a state denies parole, it uses AI to predict a likelihood of recidivism. When a digital platform returns search results or curates a news feed, it uses AI to predict, based on previous searches and other data, what users (and people like them) want to see. Scholars have identified many risks associated with these predictive uses of AI, including bias, errors, and discrimination. This talk focuses on the pathologies of prediction itself, which remain understudied. Its goal is to determine when, if ever, historic information can be a useful signal that should anchor predictions or decisions about a current problem, and when historic information should be cast aside as an unfair or inappropriate basis for decision-making. About the Speaker: Ari Ezra Waldman is a professor of law and sociology at the University of California, Irvine. His research focuses on the effects of technology on society and on queer populations in particular. He holds a PhD in sociology from Columbia University and a JD from Harvard College.