This talk argues that causal knowledge—not associative pattern recognition—is essential for achieving true fairness in algorithmic decision-making. Dominant machine learning models rely on correlations in observational data, but fairness requires understanding how interventions (e.g., policy changes) alter outcomes across socially salient groups. I demonstrate that racial fairness is a necessary condition for general algorithmic fairness, and causal inference is indispensable for diagnosing disparities, assigning responsibility, and designing effective interventions. I address key objections (e.g., challenges modeling socially constructed variables like race) and propose solutions, emphasizing that without causal frameworks, algorithms risk reinforcing systemic inequities.

Alexander Williams Tolbert is an Assistant Professor of Data and Decision Sciences at Emory University, with secondary appointments in Philosophy, Computer Science, and African American Studies. He is a Faculty Fellow at the Emory Center for Ethics and a Faculty Affiliate at the Emory Center for Mind, Brain, and Culture. His research spans causal inference, algorithmic fairness, philosophy of race, and AI ethics. He holds a PhD in Philosophy and an MA in Statistics from the University of Pennsylvania. His recent work includes publications in Synthese, Philosophy of Science, and proceedings of the ACM FAccT Conference, focusing on reconciling causal methodology with challenges of racial equity in algorithmic systems.