Bits of individual knowledge in bytes of machines

Designing policy interventions involve assuming that certain information is reliably given about individuals in a society. This assumption has remained at the center of ongoing dispute between behavioralists and rationalists over theorizing and modeling human behavior. Relatedly, the impossibility of preserving individual knowledge in integrated forms posed by Hayek is still viewed as an unresolvable knowledge problem for policymakers by many scholars. Observing that nowadays, both policytakers and policymakers are augmented agents, we revisit the knowledge problem and assumption of given information through the lens of AI. We sketch building blocks for a theory that configures and predicts the behavior of humans who are, in every step of making decisions, augmented with machines’ computational power. The resulting framework does not resolve these theoretic disputes but leapfrogs them, and recasts the practice of policy design in a new light.