Michigan Tests Prison Scoring System on Dec. 1, Targeting 30% Better Violence Prediction
Updated
Updated · Fox News · Nov 10
Michigan Tests Prison Scoring System on Dec. 1, Targeting 30% Better Violence Prediction
3 articles · Updated · Fox News · Nov 10
Summary
Dec. 1 is the start date for Michigan’s pilot of a new prison security classification system at selected men’s facilities, aimed at improving placement decisions for roughly 32,000 incarcerated people.
The software does not use AI to assign security scores: Recidiviz says it applies a rules-based formula to prison records, while corrections staff can review, override and make the final classification decision.
Michigan says the redesigned process could predict future prison violence about 30% better by using 10 years of classification data, weighting recent misconduct more heavily and allowing ongoing reassessments when records change.
That claim remains thinly documented in the state’s announcement, and any score can still be skewed by bad disciplinary data or false positives that affect housing, supervision, education and job-training access.
Even a more accurate system may be constrained by capacity: Michigan is adding 870 high-security beds, but more than 400 new beds at two facilities are expected to take about 2.5 years.
If Michigan lacks high-security beds until 2029, how will this new algorithm safely house inmates identified as high-risk today?
Could a continuous behavioral scoring system actually increase prison tension by placing incarcerated individuals under relentless surveillance and pressure?
How can we trust the 30% violence reduction claim when the algorithm was only tested on historical, potentially flawed data?