COMPAS Algorithm Case Study

The Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) is a prominent example of Algorithmic Bias in the criminal justice system. This case study demonstrates the real-world impact of biased AI systems on people’s lives.

Overview

COMPAS is a software tool developed by Northpointe (now Equivant) that:

  • Assesses likelihood of defendant reoffending
  • Used in various U.S. jurisdictions (New York, Wisconsin, California)
  • Aids courts in decisions regarding:
    • Bail
    • Sentencing
    • Parole

Bias Discovery

A 2016 ProPublica investigation revealed significant racial bias:

  • Disproportionately labeled Black defendants as high-risk
  • Higher false positive rates for Black defendants
  • Lower accuracy in predictions for minority groups

Impact on Justice System

The COMPAS case highlighted several critical issues:

Key Findings

  1. Predictive Accuracy:

    • No more accurate than non-experts
    • Demonstrated systematic bias against minorities
    • Failed to achieve intended objective of fair risk assessment
  2. Bias Manifestation:

    • Black defendants twice as likely to be misclassified as high-risk
    • White defendants more likely to be misclassified as low-risk
    • Perpetuated existing systemic biases

Lessons Learned

  1. Importance of Auditing:

    • Regular evaluation of AI systems
    • Independent third-party assessments
    • Transparency in algorithmic decision-making
  2. Need for Safeguards:

  3. Policy Implications:

    • Regulation of AI in sensitive domains
    • Standards for algorithmic fairness
    • Protection of individual rights

Learn more about the COMPAS algorithm and its implications