Many scholars have wrestled with what I call the “first-order question” in patent law: What policies should we adopt to promote innovation? This Article grapples with the second-order question: What policies should we adopt to promote innovation about promoting innovation? I argue that empirical progress in patent law depends on greater policy diversity (rather than the current emphasis on uniformity), but unconstrained “laboratories of experimentation” are suboptimal due to the spillovers from local policies. Instead, patent policy makers should adopt a third way between uniformity and local control: centralized promotion of policy variation. The optimal approach to such policy experimentation depends on the context. Randomized policy experiments should be used more often, both in the field (for example, testing prizes in a random selection of pharmaceutical classes) and in the lab (for example, testing how varying disclosure affects performance in implementing software patents).
But many nuanced, dynamic issues—such as the patent-eligibility of new technologies in heterogeneous jurisdictions—are better approached not through fixed experiments, but rather through an adaptive “experimentalist” governance regime. Local actors—patent examiners, judges, or even individual countries—should be granted broad discretion to meet centrally-defined framework goals, with the requirement of defending their decisions through robust peer review. Even where controlled experiments are infeasible, experimentalist policies could elicit local knowledge, generate varied observational data, and encourage more robust theory development about the mechanisms by which innovation policies work. This pluralistic, evidence-based approach to patent policy can be guided by recent trends in personalized and evidence-based medicine, and the resulting framework for legal experimentation has implications for policy learning beyond patent law.