# Fair prediction with disparate impact: A study of bias in recidivism prediction instruments

## TL;DR

This paper examines bias in recidivism prediction tools, showing that fairness criteria cannot all be met when group prevalence differs, and highlights how error rate imbalance leads to disparate impact.

## Contribution

It analyzes the limitations of current fairness criteria in recidivism prediction and demonstrates how error rate imbalance causes discrimination when group differences exist.

## Key findings

- Fairness criteria cannot all be satisfied with differing group prevalences.
- Error rate imbalance can lead to disparate impact.
- Current fairness measures may be insufficient to prevent bias.

## Abstract

Recidivism prediction instruments (RPI's) provide decision makers with an assessment of the likelihood that a criminal defendant will reoffend at a future point in time. While such instruments are gaining increasing popularity across the country, their use is attracting tremendous controversy. Much of the controversy concerns potential discriminatory bias in the risk assessments that are produced. This paper discusses several fairness criteria that have recently been applied to assess the fairness of recidivism prediction instruments. We demonstrate that the criteria cannot all be simultaneously satisfied when recidivism prevalence differs across groups. We then show how disparate impact can arise when a recidivism prediction instrument fails to satisfy the criterion of error rate balance.