Belinda Zou Li

Li develops ways for AI systems to form models of the world, their users, and their own internal behavior that people can inspect and update. Her work uses model introspection and natural-language explanations to make advanced systems more predictable, reliable, and useful as collaborators.