Criminal legal systems around the world are increasingly relying on software output to convict and incarcerate people. In a growing number of cases each year, governments makes these consequential decisions based on evidence from ML and statistical software that defense counsel cannot fully cross-examine. Responding to this need, we are developing audit frameworks and conducting validity studies of widely-used evidentiary statistical software (ESS) to confront the problem of wrongful convictions. Our work draws on scholarship in robust machine learning, algorithmic fairness, and the law, as well as insights from our collaborations with public defender and non-profit organizations.
Please read our recent paper for an overview:
Abebe, Rediet, Moritz Hardt, Angela Jin, John Miller, Ludwig Schmidt, and Rebecca Wexler. "Adversarial Scrutiny of Evidentiary Statistical Software." In 2022 ACM Conference on Fairness, Accountability, and Transparency, pp. 1733-1746. 2022. (Link)
If you are interested in learning more or joining our activities including upcoming workshops, database releases, ongoing research, and more, please fill out this form.