What's Happening?
The Michigan Department of Corrections is set to pilot a new scoring system on December 1 at selected men's prisons to determine security levels for incarcerated people. This system aims to improve predictions of future prison violence by approximately
30%. Contrary to some headlines, the nonprofit developing the software, Recidiviz, clarifies that the system does not use artificial intelligence to calculate security scores. Instead, it employs a points-based formula based on predetermined rules and historical Michigan prison classification data, with corrections staff making the final decisions. The new system will allow for ongoing reviews of security levels using updated prison records, rather than scheduled intervals, potentially recognizing behavioral changes sooner. Factors like recent disciplinary incidents and completion of rehabilitation programs will influence scores, with staff retaining the ability to override scores based on individual circumstances like medical needs or safety concerns.
Why It's Important?
The implementation of this new scoring system in Michigan prisons is important because it seeks to enhance safety for both correctional officers and incarcerated people by more accurately classifying individuals based on their current behavior and needs. Security classifications significantly impact an incarcerated person's daily life, affecting their housing, supervision levels, and access to educational and vocational programs. A more dynamic and responsive system could lead to fairer classifications and potentially reduce violence. However, the reliance on prison records means that inaccurate information could influence scores, raising concerns about fairness and the potential for misclassification. The distinction between a rules-based algorithm and AI is crucial, as it highlights the level of human oversight and the transparency of the decision-making process, which can impact trust in the system.
What's Next?
The pilot program for the new scoring system will begin on December 1 at selected men's prisons in Michigan. If successful, the technology could be expanded to additional facilities. The Michigan Department of Corrections expects the redesigned system to improve its ability to predict future violence by approximately 30%, though the methodology for this calculation has not been fully detailed. A key challenge identified is the current shortage of higher-security beds within Michigan's prison system. Even if the software accurately identifies individuals requiring tighter supervision, the lack of appropriate housing could hinder effective implementation. The state has announced plans to add 870 high-security beds, with some facility conversions underway and larger construction projects expected to take about two and a half years. This highlights the need for infrastructure and staffing to match technological advancements.
Beyond the Headlines
The broader implications of this initiative touch upon the intersection of technology, justice, and human rights. While the system aims to improve safety and fairness, the potential for algorithmic bias, even in a rules-based system, remains a concern if the underlying data or rules reflect existing systemic inequalities. The debate over whether such critical decisions should be made by computer-generated scores or remain entirely in human hands underscores ethical considerations regarding accountability and the human element in correctional management. Furthermore, the issue of resource allocation, specifically the availability of appropriate housing and staffing, reveals a systemic challenge that technology alone cannot solve. The success of this system will ultimately depend not just on its predictive accuracy, but also on the Michigan Department of Corrections' ability to provide the necessary resources and human oversight to act on its insights effectively and equitably.
