Imperfect Oracle: What AI Can and Cannot Do

Professor Cass R. Sunstein, Harvard Law School

Free webinar hosted by the International Telecommunications Society, the Henry N.R. Jackman Faculty of Law, and TELUS.

Human judgment is remarkable but imperfect. People are biased, and their decisions are noisy: the same problem can produce different answers depending on who is deciding, when, and under what circumstances. AI offers genuine progress on both fronts. It can evaluate more evidence, apply criteria more consistently, identify patterns people miss, and help institutions make better decisions at scale.

But the better AI becomes at judgment, the more important it becomes to understand what it cannot know.

That tension lies at the heart of Professor Cass Sunstein’s recent book, Imperfect Oracle: What AI Can and Cannot Do. Greater predictive power does not make the future fully knowable. Some outcomes depend on information that does not yet exist, complex causal interactions, changing preferences, human creativity, and events that cannot reliably be inferred from the past.

Those limits matter for institutions adopting AI. When should governments, businesses, regulators, and other organizations rely on AI to improve judgment? Where does uncertainty remain fundamental? And even where AI can predict accurately, which decisions still require human accountability because they involve values, trade-offs, and public choice?

Professor Sunstein will explore both the promise and limits of AI-assisted judgment—and what they mean for institutions trying to use increasingly capable systems responsibly and effectively.

For Canadian lawyers, the session will qualify for one hour of professional development credit.