Skip to main content
Incident intelligence/SS-IR-003CASE FILE OPEN
Symbolic editorial illustration for SS-IR-003SERVANTSTACK // INCIDENT INTELLIGENCEFORENSIC IMAGE // VERIFIED FRAME
SS-IR-003 // INCIDENT REPORTOfficial finding

COMPAS Algorithm

Racial Bias in Criminal Sentencing

EXECUTIVE BRIEF

Courts across the United States adopted COMPAS (Correctional Offender Management Profiling for Alternative Sanctions), an AI system that predicts recidivism risk to guide sentencing and bail decisions.

SURVEILLANCE & POLICING
FAILURE CHAINTRACE COMPLETE
  1. 01TRIGGERCourts across the United States adopted COMPAS (Correctional Offender Management Profiling for Alternative Sanctions),…
  2. 02MACHINE ACTIONDecision system
  3. 03MISSING GATENamed SME review and decision audit trail
  4. 04IMPACTRights & due process
01 // INCIDENT SUMMARY

The short version

Courts across the United States adopted COMPAS (Correctional Offender Management Profiling for Alternative Sanctions), an AI system that predicts recidivism risk to guide sentencing and bail decisions.

02 // KEY FACTS

Case telemetry

INCIDENT
SS-IR-003
DATE
2013-Present
SYSTEM
COMPAS Algorithm
LOCATION / SCOPE
United States
EVIDENCE
Official finding
AI ROLE
Decision system
HARM
Rights & due process
SOURCES
1 cited record
03ENTRY POINT // WHAT HAPPENED

The event

Courts across the United States adopted COMPAS (Correctional Offender Management Profiling for Alternative Sanctions), an AI system that predicts recidivism risk to guide sentencing and bail decisions. A ProPublica investigation found the algorithm was twice as likely to falsely label Black defendants as high-risk compared to white defendants, while being twice as likely to falsely label white defendants as low-risk.

04CAUSAL TRACE // AI'S ACTUAL ROLE

What the machine did

COMPAS used 137 features to generate risk scores. Judges used these scores - often without understanding the underlying model - to make bail, sentencing, and parole decisions. The algorithm's inputs correlated with race through proxy variables like zip code, employment history, and family criminal history. No human SME reviewed the model for bias before deployment to courtrooms.

Decision systemAutomation was a causal participant—not a decorative label for the system around it.
05BLAST RADIUS // CONSEQUENCES

Where the failure landed

Thousands of defendants received harsher sentences based on a biased algorithm. Black defendants who did not reoffend were flagged as high-risk at nearly twice the rate of white defendants. The system remains in use in multiple jurisdictions despite documented bias.

06 // EVIDENCE STATUS

Official finding

Supported by a court, regulator, inquiry, or other official record cited below.

SOURCE RECORD UPDATED 2026-07-09

07 // SOURCE LEDGER

1 cited record

  1. 01
08CONTROL FAILURE // MISSING GOVERNANCE

Named SME review and decision audit trail

The failure pattern in this case: Automated judgment without accountable review.

09INTERVENTION POINT // HUMAN IN THE MIDDLE

The moment the path could change

A qualified reviewer tests the basis, context, and disparate impact before the decision reaches a person.

AI PROPOSES→HUMAN OWNS THE DECISION→SYSTEM EXECUTES
10CONTROL DEPLOYMENT // AUTHORITYGATE

SME routing · decision audit trail

AuthorityGate's framework requires bias auditing by domain SMEs before any AI system is deployed in high-stakes decisions. A criminal justice expert reviewing COMPAS's feature set would have flagged zip code and family history as racial proxies. The framework also mandates ongoing fairness monitoring - not just pre-deployment testing - with human review of outcomes across demographic groups.

RELEVANT KEYSTONE CONTROLHuman-in-the-Loop ValidationHow high-risk actions route to a named subject-matter expert who owns the go or no-go decision.
12 // THE ALTERNATIVE

Autonomy is a design choice.

See the operating model that keeps AI useful while preserving human authority at consequential moments.

Compare AgenticAI and AugmentedAI →