Is legal cognition computational? (When will DeepVehicle replace Judge Hercules?)

Abstract

Could we insert machine learning into the adjudicative process? This chapter considers the extent of the isomorphism between common-law reasoning from prior cases and machine learning reasoning from prior observations, as well as the normative considerations governing any such use. It ultimately concludes that we could use machine learning models to assist judges in reasoning about some questions of law, but only in the context of an ordinary legal process regulating both the use and the forms of such models. Ultimately, machine learning would be less likely to replace judicial reasoning and legal argument than to move it around.