Legal Data: Bias in the Law, and How Legal Technology Can Be Built to Help Correct For It

Abstract

There is significant racial disparity in U.S. incarceration rates, with African Americans vastly overrepresented in prisons and jails. This was true a century and a half ago, and it is true today. Surprisingly, and in spite of decades of social psychological research, we still do not have clear answers to fundamental questions undergirding this fact: what accounts for the racially disproportionate rates of incarceration, and at what points in the criminal justice process does bias emerge, causing minority defendants to be treated differently than White defendants?In this dissertation, I cover three aspects of this problem. First, I consider bias in the law, focusing on the criminal justice actors most central to plea bargaining: criminal defense attorneys and prosecutors. I also focus on their decision making during a critical period: post-arrest and pretrial. Second, I consider stereotypes of criminal subtypes and what these stereotypes mean for legal decision making. Third, I propose that the most feasible and promising approach is to guide criminal justice actors’ decision making through the use of machine-generated outputs. Across theoretical and empirical work, I outline the case for this approach and a potential psychological difficulty that may stand in the way of making it a reality.