Which Demographic Features Are Relevant for Individual Fairness Evaluation of U.S. Recidivism Risk Assessment Tools?

TL;DR

This paper investigates which demographic features should be used in assessing the individual fairness of U.S. recidivism risk tools, finding age and sex relevant but race should be ignored.

Contribution

It provides empirical evidence on relevant demographic features for individual fairness in recidivism risk assessments, addressing a gap in legal and technical standards.

Findings

Age and sex are relevant features for fairness evaluation.

Race should be ignored in the similarity function.

Supports aligning technical fairness with legal principles.

Abstract

Despite its constitutional relevance, the technical ``individual fairness'' criterion has not been operationalized in U.S. state or federal statutes/regulations. We conduct a human subjects experiment to address this gap, evaluating which demographic features are relevant for individual fairness evaluation of recidivism risk assessment (RRA) tools. Our analyses conclude that the individual similarity function should consider age and sex, but it should ignore race.