Learning the Law Together: Judges, Litigants, and Case-by-Case Adjudication

We examine a model of case-by-case learning by judges and litigants. A judge hearing cases learns partial information about the best legal rule, gradually partitioning the case space. The evolution of doctrine is path dependent but displays strong limit properties, converging to the best legal rule. Litigant behavior strongly affects the speed of convergence. If existing case law induces litigants to modify their behavior, convergence is faster because more cases bring new information. Also, if processing information about cases is costly, the judge will optimally stop learning before convergence, leaving residual uncertainty in the law and some cases wrongly decided.