In Michigan prisons, security classifications determine access to job training and education. Now, Michigan is testing a computer-generated scoring system to influence whether an inmate stays in a low-security dorm or a high-security cell. State officials claim this can improve violence prediction by about 30%. This raises stakes for prisoner and staff safety.
However, reports clarified that artificial intelligence doesn’t assign inmates a danger score. Instead, the system follows rules, leaving the final call to staff. This prompts questions on its effectiveness.
Why Security Level Matters
The Michigan Department of Corrections manages approximately 32,000 inmates. They use manual scores to dictate where an inmate stays and monitors their activities. These scores also affect access to educational and vocational opportunities. For long-term inmates, security level changes can significantly impact daily life.
Michigan categorizes inmates into four general population levels: I, II, IV, and V. Level I allows more freedom while Level V imposes strict supervision. Lower-level inmates often access more education and job training.
Upcoming Changes with Scoring System
On December 1, selected men’s prisons will pilot the new scoring system. Success could lead to broader implementation. This system will allow real-time security level assessment using updated inmate records and reassessments based on recent behaviors or completed rehab programs.
Developed by Recidiviz, a nonprofit technology group, the tool utilizes historical prison classification data. It assigns scores based on recent behavior, allowing more accurate risk classification. Corrections staff can override scores for those with special needs, ensuring more suitable housing assignments.
AI and the New System
Investigations reveal no AI use in Michigan’s new system. Instead, Recidiviz employs a formula—a set of computer instructions that add up points based on data. Though AI aids in other corrections areas, its role here is minimal. Still, using a straightforward scoring system doesn’t ensure error-free results because inaccurate data can skew scores.
Challenges and Solutions
Predicting prison violence with 30% improvement is bold, but details on this validation are scant. Concerns include the system’s accuracy and if corrections officers agree with classifications made by these algorithms.
Facility space poses another challenge, as raised by Byron Osborn, head of Michigan corrections officers union. A lack of high-security spaces complicates housing placements within higher risk assessments.
Michigan plans to create 870 high-security beds through the Safe Prisons Initiative. This includes new units at Oaks and St. Louis Correctional Facilities. However, projects require time, taking about 2.5 years to complete. Immediate solutions remain necessary.
Increased violence incidents in Michigan prisons highlight the urgency. Improved classification could foresee dangerous situations earlier, but adequate resources and staffing are vital for impactful action.
“Classifying prisoners at the security level that fits their current behaviors and needs keeps everyone safer,” Michigan Corrections Director Heidi Washington.
Conclusion
Michigan’s initiative to enhance prison safety is commendable. However, reliance on traditional computing over AI points to cautious progression. While security scores may evolve, success hinges on both technological and infrastructural investments.

Trump Defends Time on White House Improvements Amid FAA Issues
President Trump Addresses Reporters Amid Controversy
FBI Leads Investigation into Bomb Threat at Ethanol Facility in Peoria
Tanker Steering Malfunction Causes Collision with Cruise Ship in New Orleans
Longest Ground Stoppage Since 9/11 Ends After Technical Glitch
Nvidia Founder Jensen Huang’s Potential Return Backed by Tech Leaders