UN Innovation Network 27d ago

This session with Sanmi Koyejo (Stanford University) explores how AI systems can be evaluated, tested and adapted across diverse global contexts. Drawing on themes from AI evaluation across different settings, human preferences in AI, and his work in Africa, the session considers what it takes to build systems that are useful, reliable and responsive to different communities. The conversation examines how AI performs across contexts, and what this means for fairness, trust, accountability and practical use. It also explores how behavioural science can help institutions understand how people adopt, rely on and respond to AI in practice.