Events

How can evaluation help us understand AI systems, improve their design, and support responsible deployment? This talk will present an overview of my research on the methodological and practical challenges of evaluating AI. Through selected examples, I will examine how data, assumptions, metrics, and context shape what evaluations reveal—and how those findings can inform system design. The talk will span causal and statistical foundations as well as the assessment of AI systems in practice, concluding with a research agenda for connecting rigorous evaluation to better decisions across the AI lifecycle.