Association for Public Policy Analysis & Management

Few policy areas have seen such intense ethical scrutiny in recent public discussions as machine learning applications in the criminal justice system. In particular, applications that are used to make pre-trial detention decisions, such as Equivant’s Northepointe Suite, have been criticized as being unfair against people of color (Angwin et al. 2016).

In this paper, I use this case of pre-trial detention as a case study for investigations that aim to make machine learning applications for public services fair. I distinguish competing technical definitions of fairness, review limitations to achieving such fairness in practice (Lum 2017; Eckhouse et al. 2018), and in principle (Kleinberg, Mullainathan, and Raghavan 2016; Corbett-Davies and Goel 2018; Berk et al. 2017). Building on this existing work, I identify central dilemmas for policy makers, such as trade-offs between different competing notions of technical fairness. I contrast technical approaches to fairness with approach to fairness developed from the normative perspective of moral and political philosophy. From this case study of criminal justice, I illustrate lessons as they apply for the management and regulation of machine learning applications in public services more broadly.

References

Angwin, Julia, Jeff Larson, Surya Mattu, and Lauren Kirchner. 2016. “Machine Bias.” Text/html. ProPublica. May 23, 2016. https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing.

Berk, Richard, Hoda Heidari, Shahin Jabbari, Michael Kearns, and Aaron Roth. 2017. “Fairness in Criminal Justice Risk Assessments: The State of the Art.” ArXiv:1703.09207 [Stat], March. http://arxiv.org/abs/1703.09207.

Corbett-Davies, Sam, and Sharad Goel. 2018. “The Measure and Mismeasure of Fairness: A Critical Review of Fair Machine Learning,” September, 25.

Eckhouse, Laurel, Kristian Lum, Cynthia Conti-Cook, and Julie Ciccolini. 2018. “Layers of Bias: A Unified Approach for Understanding Problems With Risk Assessment.” Criminal Justice and Behavior, November, 0093854818811379. https://doi.org/10.1177/0093854818811379.

Kleinberg, Jon, Sendhil Mullainathan, and Manish Raghavan. 2016. “Inherent Trade-Offs in the Fair Determination of Risk Scores.” ArXiv:1609.05807 [Cs, Stat], September. http://arxiv.org/abs/1609.05807.

Lum, Kristian. 2017. “Limitations of Mitigating Judicial Bias with Machine Learning.” Nature Human Behaviour1 (7): 0141. https://doi.org/10.1038/s41562-017-0141.