Michigan Tests New System for Prison Security Classification

Incarcerated men in Michigan have participated in employment counseling within the state’s prisons. The security classifications they receive impact their access to educational and job training opportunities (Michigan Department of Corrections).

The Michigan Department of Corrections is pioneering a new system that could influence whether inmates reside in a low-security dorm or a more controlled facility. The state claims the system may enhance predictions of potential prison violence by about 30%, a significant assertion when the well-being of both correctional officers and inmates depends on accurate assessments.

Understanding the System

The headlines may suggest that artificial intelligence determines the danger level of inmates, yet the software’s developers clarify that AI is not involved in calculating these security scores. The system adheres to a set of established rules, with final decisions made by corrections staff. This technology does not replace human judgment but instead supports it.

Security Levels and Daily Life

Michigan’s Department of Corrections is responsible for the safety and oversight of nearly 32,000 incarcerated individuals. The manual process currently used to assign security levels affects both living conditions and access to programs.

In Michigan, there are four general population security levels: I, II, IV, and V. Level I, with more freedom, contrasts with Level V, which is tightly supervised. Lower levels often provide increased access to educational and vocational programs.

Revamping the Classification System

On December 1, Michigan plans to pilot a new classification approach. Unlike the current system, it allows for ongoing updates based on current behavior and needs, using prison records to reassess security levels in real-time.

Developed by nonprofit technology organization Recidiviz, the program uses a decade of data to revamp the scoring process. It evaluates prisoner records and calculates scores based on criteria evaluated by corrections experts. Recent incidents are more heavily weighted, while consistent good behavior could reduce restrictions.

The new scoring system also considers rehabilitation efforts. The software flags records when new data suggests a review is required. This approach aims to improve safety by ensuring that security classifications reflect current behavior and needs.

Potential Challenges

Despite optimism about the system’s potential, some skepticism remains. Recidiviz uses AI for other aspects of its platform, but not in assigning security levels. This distinction underscores the need for transparency, as computerized systems must be accurate and reliable.

Michigan’s claim of a 30% improvement in predicting violence is significant, yet the specifics behind this claim remain unclear. Questions arise about how the system will balance accurate predictions with avoiding misclassification of inmates.

Practical Considerations

The accuracy of the new classification system interacts with broader logistical issues. Michigan faces a shortage of high-security beds, creating challenges even if a security concern is identified. Solutions include plans to add 870 high-security beds as part of Michigan’s Safe Prisons Initiative.

Until those facilities are complete, challenges will persist. Recent in-prison violence underscores the need for effective classification, sufficient staffing, and adequate housing. Enhancements in classification may identify risks, but addressing them necessitates comprehensive resources.

Ultimately, Michigan’s move towards technology-driven solutions seeks to create safer prison environments. As of December 1, the pilot may set a precedent for change, aligning with broader goals of safety and fairness within the correctional system.