Henrik Sigstad (University of Oslo); Daniel L. Chen (Toulouse School of Economics)
Economics / Institutions and organizations in the public sector
Abstract
What is the impact of artificial intelligence on the legal system? In a general deterrence model, we show that although machine learning might optimally reduce type I and type II errors, basing legal decisions on machine predictions can undermine incentives to abide by the law. We discuss under which assumptions machine learning methods can be adapted to obtain optimal deterrence. In a planned empirical application on a corpus of 14 million Brazilian labor lawsuits, we assess the amount of statistical discrimination under various machine learning decision rules.