Regulation by Machine

Abstract

Legal scholars investigating artificial intelligence are preoccupied with regulation. The literature has largely focused on the need for humans to regulate the behavior of automated systems. In this paper, we focus on the converse: how artificially intelligent systems can serve to regulate human behavior. The shortcomings of human-led regulation are clear. We argue that machine learning technology can address some of these limitations. We provide examples of how machine learning can predict how courts would decide legal disputes more cheaply and accurately than human regulators. This allows regulators to streamline operations, providing fast, accurate, consistent, and reliable ex ante regulatory advice and rulings. We further explore how machine learning technology might soon be used to refine laws and reduce errors.

Keywords: machine learning, artificial intelligence, regulation

JEL Classification: K00, K10, K20, K23