A Framework for the Efficient and Ethical Use of Artificial Intelligence in the Criminal Justice System

Machine learning techniques are transforming the manner in whichmuch of the legal system works, and criminal justice is the area whichwill be most fundamentally changed. Given the fundamental rightsand interests at stake in the criminal justice system, this is also thefield where the unthinking application of artificial intelligence ("AI") ismost troubling, and where there is the greatest threat to individualrights and the likelihood of unanticipated damage to the rule of law.These problems will occur (and are occurring) throughout thecriminal justice system: from data-driven predictive policing systemsin the criminal investigation process, through to recidivism predictionfor parole applications and sentencing recommendation systemspost-trial. The risks presented by Al to the proper functioning of thecriminal justice system will be exacerbated by commercial pressureson law enforcement and the criminal justice system, partisan politicalinterests, and a lack of technological understanding by the judiciaryand the legal profession more generally. Notwithstanding thisdystopian vision, there is an opportunity to use AI techniques toimprove the detection of crime, prosecute and sentence criminaloffenders, help uncover discrimination, ensure parity of treatmentacross the system, and identify unfair and unjust treatment. Thethoughtful and appropriate use of "ethical" Al systems can greatlyassist in the administration of justice and the rule of law. In thisArticle, we propose a framework for systematically implementingAl into the criminal justice system in order to ensure that the systemoperates in a normatively enhanced and more effective and efficientmanner. In proposing this framework we grapple with the realitythat humans have an intrinsic emotional dislike of computers makingdecisions that have an important impact on peoples' lives