Risk assessment algorithm PSA, used to see if pretrial defendants in NJ should be held in jail, has led to jail population drop, yet racial disparities persist
A nonprofit group encouraged states to use mathematical formulas to try to eliminate racial inequities. Now, it says the tools have no place in criminal justice. Tweets: @demandprogress , @wired , and @wired Tweets: @demandprogress : Algorithms are often racist because they are created by people. They have no place in the criminal justice system because they perpetuate racism — and they have no place in determining “kill lists” by the government in the name of national security. https://www.wired.com/... @wired : The Pretrial Justice Institute encouraged states to use algorithms to try to eliminate racial inequities. Now, the group says the risk assessment tools it previously promoted have no place in pretrial justice because they perpetuate racial inequities. https://www.wired.com/... @wired : If you are booked in New Jersey, a judge will decide whether to hold you until trial or set you free. One factor the judge must weigh: the result from an algorithm called PSA. Now the group behind the algorithm says it has no place in criminal justice. https://www.wired.com/...
Context & Ripple Effects
The Pretrial Justice Institute built its reputation urging states to swap cash bail for actuarial tools, and New Jersey's PSA became the flagship result: the jail population fell. But the same institute now says tools like PSA have no place in pretrial justice, a reversal that lands six years after [[a:869915|ProPublica's analysis found strong bias against black defendants in crime-prediction software]] used in US sentencing.
The timing matters: states had been legislating along the institute's original playbook — California replaced cash bail with an algorithmic risk system in 2018 — and coverage of predictive tools spreading across policing, sentencing, and probation shows how deeply the approach had been institutionalized before the promoter turned against it.
First-order effects
- New Jersey's pretrial system keeps the PSA tool's headline achievement — a smaller jail population — but must now defend a process whose own promoter says perpetuates racial inequity among defendants.
- The Pretrial Justice Institute's about-face strips the reform movement of its most credible institutional endorsement for algorithmic risk assessment, leaving states that adopted PSA-style tools citing that endorsement exposed.
Second-order effects
- States that legislated algorithmic bail on the institute's advice, most visibly California's cash-bail replacement, face pressure to reassess whether the fix reproduces the disparity it was sold as removing.
- Vendors of proprietary pretrial and recidivism algorithms — already operating with limited transparency as courts weigh their use — lose a key reform-community ally and gain harder scrutiny over how scores are built and validated.
Third-order effects
- If the pattern holds, criminal-justice reform splits into two camps — algorithm-as-reform versus algorithm-as-problem — and adoption decisions shift from nonprofits' endorsements toward evidence of how tools actually perform, a shift already visible in findings that judges use risk assessments selectively rather than wholesale.
- The broader legitimacy of third-party risk-scoring in courts, contested since justices began weighing the consequences of proprietary recidivism algorithms, now hinges on whether racial disparities can be eliminated at all — or whether pretrial decision-making returns to human judgment with algorithms demoted to advisory inputs.
The trend: Criminal-justice reform is pivoting from treating algorithmic risk assessment as the cure for bail inequity to treating the algorithms themselves as a source of it, with the tools' original promoters leading the retreat.