Wednesday, September 18 at 12:00pm to 1:00pm
Law Building (LAW), 3500
401 East Peltason Drive Irvine, CA 92697-8000
Richard Re, Professor of Law, UCLA School of Law
criminal procedure, federal courts, constitutional law
A Contractualist Approach to Algorithmic Criminal Justice
This paper contributes to debates on algorithmic fairness in the criminal justice system by drawing on contractualism, an ethical theory that demands principles of conduct that no individual can reasonably reject. In general, contractualism privileges the objections of the most vulnerable individual above aggregate social welfare. Contractualism helps to explain why algorithmic methods that treat individuals differently on the basis of involuntary traits pose special fairness challenges: in general, individuals have less ability to raise reasonable objections when they could have reduced their risk of suffering burdens by choosing appropriately. Further, an individual has good reason to reject principles that would burden her based on the actions of others, particularly when doing so would compound background social inequalities. Examples include not just racial profiling but also the many predictive policing tools that rest on statistical patterns observable at a neighborhood level, without considering individual behavior. Still, contractualism suggests that algorithmic and other forms of criminal-justice prediction can be fair. While an algorithm’s “explainability” is less important to fairness than sometimes suggested, many predictive tools are at least somewhat sensitive to individual choice. Contractualism thus offers a moderate approach, intermediate between views that wholly embrace or repudiate algorithmic criminal justice tools. And contractualism’s insights can be helpful even for those who subscribe to different first-principles views of criminal justice.
Lunch will be provided.
Intellectual Life Workshop Calendar
You may login with your UCINetID. No account creation is necessary
ALL OTHER USERS:
When creating an account, please create it with your first and last name, otherwise it may be deleted
NOTE ON EVENT SUBMISSION:
In order to submit events, users must have an UCI email address (@uci.edu). All events submitted by non-UCI users WILL be deleted.