Learn how to bridge prediction and intervention problems in social systems through this 29-minute conference talk by Ashia Wilson from MIT. Explore the fundamental challenges of designing evaluations that effectively connect predictive modeling with real-world interventions in complex social contexts. Discover methodological approaches for addressing the gap between what machine learning models predict and how interventions actually perform when implemented in social systems. Examine case studies and theoretical frameworks that demonstrate how to design robust evaluation strategies that account for the unique characteristics of social environments, including feedback loops, behavioral responses, and systemic interactions that can affect intervention outcomes.