Prediction, Performativity, and Potential Outcomes - Communicative Rationality in Prediction-Allocation Problems

Explore the complex relationship between prediction systems and their real-world impacts in this 16-minute conference talk that examines how predictive models influence the very outcomes they aim to forecast. Delve into the concept of performativity in algorithmic decision-making, where predictions don't merely describe future states but actively shape them through allocation decisions and behavioral responses. Learn about the theoretical framework of potential outcomes and how it applies to prediction-allocation problems, particularly in contexts where human-AI collaboration affects decision-making processes. Understand the notion of communicative rationality as it relates to predictive systems and their role in shaping social and economic outcomes. Examine case studies and theoretical models that demonstrate how prediction systems can create feedback loops, alter incentives, and ultimately transform the environments they operate within. Gain insights into the ethical and practical implications of performative predictions, including considerations for fairness, accountability, and the unintended consequences of algorithmic interventions in human decision-making processes.