Recommendation in Equilibrium for the NYC High School Match

Learn how to design congestion-aware recommendation systems that maintain accuracy even when widely adopted, through a real-world case study of NYC's high school matching process. Explore the methodological challenges of creating individualized recommendations for over 70,000 8th graders applying to 800+ high school programs, where traditional recommendation approaches fail because they become self-defeating when many students follow the same advice. Discover how researchers collaborated with NYC Public Schools to develop an informational intervention that helps underserved students identify high-performing, nearby schools with strong individual admissions likelihood while ensuring the recommendations remain accurate in equilibrium. Examine the intersection of prediction and intervention in social systems, understanding how recommendation algorithms must account for their own impact on user behavior and system outcomes to avoid creating the very problems they aim to solve.