Judging the Algorithm: Constitutional and Human Rights Challenges to AI -Driven Risk Assessment in Criminal Sentencing

Algorithmic risk assessment tools, most notably COMPAS, are increasingly used in criminal sentencing to generate risk scores based on factors including criminal history, age, and socioeconomic background. However, studies have demonstrated significant racial bias in these tools, with Black defendants disproportionately assigned higher risk scores than white defendants with comparable histories. Using constitutional analysis and international human rights frameworks, this paper argues that current applications of these tools are constitutionally inadequate under US due process and equal protection doctrine.

Rather than calling for their outright elimination, this paper contends that heavy regulation, like the international models such as the EU AI Act, offers a more straightforward and practical path forward for the use of algorithmic risk assessment tools in criminal sentencing.