A Brief History of the Changing Roles of Case Prediction in AI and Law

Abstract

Predicting case outcomes has long played a role in research on Artificial Intelligence and Law. Actually, it has played several roles, from identifying borderline cases worthy of legal academic commentary, to providing some evidence of the reasonableness of computational models of case-based legal reasoning, to providing the raison d'être of such models, to accounting for statistically telling features beyond such models, to circumventing features altogether in favor of predicting outcomes directly from analyzing case texts. The use cases to which case prediction has been put have also evolved. This article briefly surveys this historical evolution of roles and uses from a mere research possibility to a fundamental tool in AI and Law’s kit bag of techniques.