Michigan’s new prison scoring system claims to enhance violence prediction by 30%, but the nonprofit behind it clarifies that it does not utilize AI for inmate classification.
Michigan is set to implement a new prison scoring system that could significantly impact the classification of inmates, determining whether they reside in low-security dormitories or high-security cells. State officials assert that this system could improve predictions of future prison violence by approximately 30%. This claim raises critical questions about the accuracy and reliability of such assessments, especially given the implications for the safety of both correctional officers and incarcerated individuals.
However, early reports about the system’s reliance on artificial intelligence (AI) have been misleading. The nonprofit organization Recidiviz, which is developing the software, has clarified that AI does not play a role in calculating security scores. Instead, the system operates based on predetermined rules, with corrections staff making the final decisions regarding inmate classifications.
The Michigan Department of Corrections oversees around 32,000 incarcerated individuals across the state. Currently, corrections employees utilize a manual scoring process to determine the appropriate security level for each inmate. These classifications not only dictate where an inmate lives but also influence their supervision levels and access to educational programs and job training opportunities. For inmates serving lengthy sentences, a change in security level can significantly affect their daily lives.
Michigan’s prison system categorizes inmates into four general population security levels: I, II, IV, and V. Level I allows for greater freedom of movement, while Level V involves stricter supervision. Inmates at lower security levels often have access to more vocational training and educational opportunities. However, the state maintains that programs such as Violence Prevention Programming are available at every general population security level. Notably, there is no general population Level III.
To enhance the classification process, the Michigan Department of Corrections plans to begin testing the new system on December 1 at selected men’s prisons. If the pilot is successful, officials may expand the technology to additional facilities. A key feature of the new system is the frequency of reassessments. Instead of relying on scheduled evaluations, staff will be able to conduct ongoing reviews using updated prison records, which could help identify behavioral changes more swiftly.
The software developed by Recidiviz analyzes ten years of Michigan prison classification data to redesign the state’s scoring process. It examines existing information in prison records and calculates a score based on factors identified by Michigan corrections experts. Recent disciplinary incidents are weighted more heavily than older ones, while inmates demonstrating good behavior over time may qualify for a lower security level. Additionally, completing certain rehabilitation programs can also lead to a less restrictive classification. The software can flag records for review when new information arises.
Heidi E. Washington, Director of the Michigan Department of Corrections, expressed optimism about the changes, stating, “Classifying prisoners at the security level that fits their current behaviors and needs keeps everyone safer.” She believes that combining historical data with research-based scoring could enhance security throughout Michigan’s prison system.
Despite the promising nature of the initiative, skepticism remains regarding the reliance on computer-generated scores. On November 2, Recidiviz published a statement clarifying that the classification system does not use AI to determine security scores. Instead, it employs a points-based formula that corrections staff can review. This formula, often referred to as an algorithm, is essentially a set of instructions that a computer follows to complete a task, akin to a digital scorecard. The same information will yield the same score, allowing staff to examine and override scores when necessary.
While Recidiviz does incorporate AI in other aspects of its technology platform, such as assisting corrections employees in preparing case notes, it does not influence security classifications or release decisions. This distinction is crucial, as it underscores the importance of human oversight in decisions that affect inmate safety and well-being.
Nevertheless, the system’s effectiveness hinges on the accuracy of the underlying data. If a disciplinary report contains inaccuracies, it could skew an inmate’s score. Moreover, risk assessments cannot predict future behavior with absolute certainty. The Michigan Department of Corrections has claimed that the redesigned system will improve violence prediction by 30%, a figure that lacks transparency regarding its calculation and underlying performance metrics.
Questions remain about the system’s ability to accurately identify potential threats and whether it might incorrectly classify individuals as dangerous. Furthermore, the system’s success is contingent upon the availability of appropriate housing for inmates classified as higher risk. Byron Osborn, president of the union representing Michigan corrections officers, has raised concerns about the shortage of higher-security beds, emphasizing that the lack of appropriate housing for individuals already classified as higher risk poses a significant challenge.
Michigan has announced plans to add 870 high-security beds as part of its Safe Prisons Initiative, with some facility conversions already underway. However, larger construction projects will take longer to complete. Osborn has warned that solutions are needed sooner, especially in light of recent violent incidents within the state’s prisons, including cases where inmates have allegedly killed their cellmates.
While better classification could help officials recognize potentially dangerous situations more quickly, the department must also address staffing shortages and housing availability to respond effectively. The question remains: how much safer can a prison become if officials identify a risk but lack the resources to address it?
As Michigan prepares to test this new scoring system, the potential for improved safety and fairer assessments is evident. However, the state must provide clarity on how it arrived at its claims of a 30% improvement and ensure that the system does not inadvertently penalize inmates who have made positive changes in their behavior. Ultimately, the success of this initiative will depend on the state’s ability to balance technological advancements with the realities of prison management.
Would you trust a computer-generated score to help determine how dangerous someone might be, or should such decisions remain entirely in human hands? Let us know your thoughts.
According to Fox News.