Decision Theoretic Foundations for Human-AI Collaboration

Learn statistical decision theory and information economics frameworks for evaluating and designing effective human-AI collaborative systems in this 31-minute conference talk from the Simons Institute. Explore measurement frameworks that address critical questions at the human-agent interface, including how to assess when decision-makers appropriately rely on model predictions and when humans or AI agents can better exploit available contextual information. Discover principled approaches to evaluating and designing explanations for AI systems, while understanding why simple methods of combining human and AI judgments can lead to apparent team performance failures despite complementary information. Examine the theoretical foundations needed to move beyond basic human-AI collaboration approaches toward more sophisticated frameworks that leverage the strengths of both human and artificial agents in decision-making contexts.