Further Reading: AI and Justice — Criminal Justice, Civil Rights, and Accountability

These sources are organized by topic and annotated to help you decide what to read next. All sources are Tier 1 (published, peer-reviewed, or from established investigative outlets) or Tier 2 (reputable institutional reports and expert analyses).

Predictive Policing

Lum, K., & Isaac, W. (2016). "To predict and serve?" Significance, 13(5), 14–19. A clear, accessible analysis of how predictive policing algorithms trained on drug arrest data in Oakland, California, reproduced racially biased policing patterns. This is the study cited in Section 17.1 — an excellent starting point for understanding the feedback loop problem. [Tier 1 — Peer-reviewed journal article]

Ferguson, A. G. (2017). The Rise of Big Data Policing: Surveillance, Race, and the Future of Law Enforcement. New York University Press. A comprehensive, readable examination of how data-driven policing technologies — including predictive policing, social network analysis, and surveillance — interact with race and civil liberties. Ferguson, a law professor, writes for a general audience while maintaining scholarly rigor. Essential reading for anyone interested in this topic. [Tier 1 — Academic monograph]

Brayne, S. (2020). Predict and Surveil: Data, Discretion, and the Future of Policing. Oxford University Press. A sociologist's account of how the Los Angeles Police Department adopted big data tools, based on years of embedded research. Brayne provides a rare inside look at how officers actually use (and sometimes resist) predictive technology. [Tier 1 — Academic monograph based on ethnographic research]

Risk Assessment and the COMPAS Debate

Angwin, J., Larson, J., Mattu, S., & Kirchner, L. (2016). "Machine Bias." ProPublica. The original investigation that sparked the national debate about COMPAS and algorithmic fairness. Clearly written, meticulously documented, and accompanied by open-source data and methodology. Start here for the narrative version of the fairness debate. [Tier 1 — Investigative journalism]

Dressel, J., & Farid, H. (2018). "The accuracy, fairness, and limits of predicting recidivism." Science Advances, 4(1), eaao5580. A peer-reviewed study showing that COMPAS was no more accurate than untrained humans making predictions based on just two variables (age and number of prior convictions). A sobering check on claims about algorithmic superiority. [Tier 1 — Peer-reviewed]

Chouldechova, A. (2017). "Fair prediction with disparate impact: A study of bias in recidivism prediction instruments." Big Data, 5(2), 153–163. The formal mathematical proof that certain fairness criteria cannot be simultaneously satisfied when base rates differ between groups. Technical but clearly written — the mathematical impossibility at the heart of the COMPAS debate. [Tier 1 — Peer-reviewed]

Kleinberg, J., Mullainathan, S., & Raghavan, M. (2016). "Inherent Trade-Offs in the Fair Determination of Risk Scores." arXiv preprint arXiv:1609.05807. Another foundational mathematical analysis of the fairness impossibility, approaching it from a slightly different angle than Chouldechova. Together, these papers establish the theoretical framework for understanding why algorithmic fairness requires human value judgments. [Tier 1 — Widely cited preprint, subsequently published in peer-reviewed proceedings]

Constitutional and Legal Issues

State v. Loomis, 881 N.W.2d 749 (Wis. 2016). The full text of the Wisconsin Supreme Court's decision in the key COMPAS case. The court's reasoning — and its explicit cautions about the use of risk assessment tools — is worth reading in full. Available free through Google Scholar or court databases. [Tier 1 — Court decision]

Selbst, A. D. (2017). "Disparate Impact in Big Data Policing." Georgia Law Review, 52, 109–195. A thorough analysis of how disparate impact doctrine (from employment discrimination law) could be applied to algorithmic policing tools. Selbst argues that existing legal frameworks can address many algorithmic harms — if courts are willing to apply them. [Tier 1 — Law review article]

Richardson, R., Schultz, J. M., & Crawford, K. (2019). "Dirty Policing: Protest and Policing Data and the Problem of Biased Data." New York University Law Review Online, 94, 15–55. Examines how the use of data from police departments with documented misconduct (consent decrees, corruption investigations) infects algorithmic systems trained on that data. An important contribution to the "dirty data" debate. [Tier 1 — Law review article]

Accountability and Reform

Raji, I. D., Smart, A., White, R. N., Mitchell, M., Gebru, T., Hutchinson, B., Smith-Loud, J., Theron, D., & Barnes, P. (2020). "Closing the AI accountability gap: Defining an end-to-end framework for internal algorithmic auditing." Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, 33–44. Proposes a practical framework for internal algorithmic auditing — useful for understanding how organizations can hold themselves accountable. [Tier 1 — Peer-reviewed conference paper]

Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019). "Fairness and Abstraction in Sociotechnical Systems." Proceedings of the Conference on Fairness, Accountability, and Transparency, 59–68. Argues that fairness cannot be solved purely through technical means — it requires attention to social context, institutional design, and political structures. A good companion to this chapter's argument that fairness metrics are political choices. [Tier 1 — Peer-reviewed conference paper]

Broader Context: Race, Technology, and Justice

Benjamin, R. (2019). Race After Technology: Abolitionist Tools for the New Jim Code. Polity Press. A provocative and influential book arguing that algorithmic systems can reproduce racial inequality under the guise of technological neutrality. Benjamin coins the term "the New Jim Code" to describe this phenomenon. Accessible and widely assigned in university courses. [Tier 1 — Academic monograph]

Eubanks, V. (2018). Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. St. Martin's Press. Examines how automated decision-making systems — in welfare, housing, and criminal justice — disproportionately affect poor and marginalized communities. Written for a general audience with compelling narrative storytelling. [Tier 1 — Trade nonfiction, extensively researched]

O'Neil, C. (2016). Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Crown. An early and influential critique of algorithmic decision-making in domains including criminal justice, hiring, and education. O'Neil, a mathematician, writes accessibly about how opaque models can reinforce existing inequalities. [Tier 1 — Trade nonfiction]

Facial Recognition and Surveillance

Buolamwini, J., & Gebru, T. (2018). "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification." Proceedings of the Conference on Fairness, Accountability, and Transparency, 77–91. The landmark study showing that commercial facial recognition systems had significantly higher error rates for darker-skinned women. Though focused on gender classification, the methodology and findings are directly relevant to facial recognition in justice contexts. [Tier 1 — Peer-reviewed]

Hill, K. (2020). "Wrongfully Accused by an Algorithm." The New York Times. The story of Robert Williams's wrongful arrest based on a faulty facial recognition match — the first publicly documented case in the United States. Clear, compelling journalism that puts a human face on algorithmic error. [Tier 1 — Investigative journalism]

Where to Go Next

• If you want to go deeper on the mathematical fairness impossibility, start with Chouldechova (2017) or Kleinberg et al. (2016).
• If you want the narrative version of the justice AI debate, start with Angwin et al. (2016) and Benjamin (2019).
• If you want to explore policy solutions, start with Raji et al. (2020) and Ferguson (2017).
• If you want to understand how policing technology works in practice, start with Brayne (2020).