RAND corporation report finds Chicago's predictive policing algorithm doesn't reduce homicides
Matt Stroud / The Verge :
Context & Ripple Effects
The RAND evaluation is the first hard audit of Chicago's person-based 'heat list,' and its null result lands in a field already under strain: critics had argued the year before that biased training data justified dismantling these systems outright, while advocates pushed for fairness fixes instead.
What makes the finding consequential is what came after it — five years later Chicago PD was still running the system, constantly surveilling one man it flagged without being able to say whether he'd be shooter or victim, while a national investigation into PredPol showed similar tools steering patrols toward poor, Black, and Latino neighborhoods.
First-order effects
- Chicago PD loses its central justification for the heat list — an independent evaluator found no homicide reduction — yet the people already flagged remain subject to heightened police attention with no demonstrated public-safety return.
- RAND's credibility as an outside auditor makes this harder for the department to dismiss than internal justifications would have been.
Second-order effects
- Vendors and departments selling predictive policing elsewhere face the same audit question: LAPD's successor effort was later shown to resemble PredPol and Operation Laser, programs already shut down amid public outcry, so a null efficacy result gives opponents a second line of attack beyond bias.
- The critique splits the market — some pressure moves toward officer-accountability analytics like Benchmark Analytics, which pitches data-driven identification of problem officers but struggles with data quality.
Third-order effects
- If independent evaluations keep returning null or biased results, person-based predictive policing faces a structural fork: dismantlement, as critics demand, or repositioning toward internal accountability tools — with the burden of proof shifting from skeptics to vendors.
- The pattern also raises the stakes for transparency norms: cities deploying opaque risk-scoring on residents can now expect third-party audits to become part of the procurement fight itself.
The trend: Independent audits are shifting the predictive-policing debate from whether algorithms can be made fair to whether they work at all, pressuring departments to justify or abandon person-based risk lists.