Towards Plurality - Foundations for Learning from Diverse Human Preferences

Explore foundational approaches to learning from diverse human preferences in this 49-minute workshop presentation by Ramya Korlakai Vinayak from the University of Wisconsin-Madison. Delve into the theoretical and practical challenges of developing systems that can accommodate plurality in human preferences rather than assuming homogeneous user needs. Examine methodological frameworks for capturing, representing, and learning from heterogeneous preference data across different populations and contexts. Discover how machine learning systems can be designed to respect and incorporate diverse viewpoints while maintaining effectiveness and fairness. Learn about the mathematical foundations underlying preference aggregation, the challenges of preference elicitation from diverse groups, and approaches to building more inclusive AI systems that acknowledge the complexity of human values and decision-making processes.