Beyond “Fairness”: Rethinking the Use of Algorithmic Predictions in Criminal Justice

Abstract

This paper critiques the widespread use of predictive algorithmic tools in criminal justice, such as COMPAS, arguing that concerns about fairness and accuracy, while important, fail to address a deeper ethical issue: the infringement of the right to be treated as an individual. Drawing on Renee Jorgensen’s work, I argue that fairness-based reforms are insufficient because predictive punishment is incompatible with the demands of negative retributivism, the most compatible theory of punishment with the demands of the right to be treated as an individual. Given the high stakes of criminal law and the inherent trade-off in algorithmic fairness metrics, I contend that algorithmic predictions should not be used to justify punishment or policing decisions. However, I propose that algorithmic tools can be ethically employed in developing policies aimed at crime reduction, provided they are used to identify causal factors rather than to predict individual behavior. To this end, I advocate a pluralist framework: negative retributivism should govern punishment and policing, while rights-based consequentialism should inform long-term policy goals. This approach aims to clarify when the use of algorithms in criminal justice is unjustified, and when it may be justified with critical revision.