Human Decision Governance

The central mistake in the AI debate, Mayer-Schönberger and Gasser argue in Guardrails: Guiding Human Decisions in the Age of AI, is that it has been conducted in the wrong register. We have debated what machines can know, whether they can be trusted to be accurate, how to reduce their errors—as though the problem were fundamentally epistemic. But the real promise and peril of AI lie not in knowledge but in decision: in the moment when a prediction or recommendation is folded into a choice, and in what happens to human agency in that moment. Three kinds of structures have always shaped human decision-making without making decisions for us: constraints on information, norms that govern how we choose among options, and consequences that bind us to our choices. These guardrails, the authors argue, are not obstacles to good decisions but their precondition—the structures that guide choice toward better outcomes while preserving