Events

CDS Colloquium: Modern Foundations of Social Prediction

Speaker: Juan Perdomo

Location: 370 Jay Street, Room 825
Videoconference link: https://nyu.zoom.us/j/98082078176

Date: Monday, March 17, 2025

Current algorithmic prediction systems do not just forecast the world around us; they actively shape it. They steer markets, influence individual opinions, and determine access to social goods. This dynamic, where we build systems using historical data to influence future behavior, underscores the role of prediction as both a lens and engine of social patterns. It also calls for developing modern foundations of machine learning that explicitly account for the feedback between algorithms and society and ensure that our new technologies have a positive impact. In this talk, I will present work tackling various core challenges in this agenda. The first part will focus on performative prediction, a learning-theoretic framework that addresses the causal aspects of social prediction. In the second half, I will discuss an empirical case study evaluating the impact of a risk prediction tool used to allocate interventions to hundreds of thousands of public school students in Wisconsin each year. I’ll close with some discussion of future work and the exciting directions that lie ahead.