How predictive algorithms are increasingly being used to set police patrols, prison sentences, and probation rules in the US and Europe
Across the United States and Europe, software is making probation decisions and predicting whether teens will commit crime. Opponents want more human oversight. Tweets: @satariano , @carnage4life , @karaswisher , @lynnnovick , @justinhendrix , @katrina_hrm , @princeton , @defensebaron , @bigbrotherwatch , and @helengymatlarge Tweets: Adam Satariano / @satariano : For months, @CadeMetz and I have been chipping away at a story about people in the U.S. and Europe whose lives have been affected by government decisions made by algorithms. It included stops in Philadelphia, Rotterdam and Bristol. We're excited to share: https://www.nytimes.com/... Dare Obasanjo / @carnage4life : It's wild that the same algorithmic techniques that predict you'd like alt-right videos screaming about SJWs because you watched a gaming video on YouTube is now being used to determine jail sentences & parole based on predicting one's likelihood to commit https://www.nytimes.com/... Kara Swisher / @karaswisher : Minority Report jumps from fiction to reality: via @NYTimes https://www.nytimes.com/... Lynn Novick / @lynnnovick : “Where is my human interaction?” Computers are not people. An Algorithm That Grants Freedom, or Takes It Away https://www.nytimes.com/... Justin Hendrix / @justinhendrix : An Algorithm That Grants Freedom, or Takes It Away: Across the United States and Europe, software is making probation decisions and predicting whether teens will commit crime. Opponents want more human oversight. https://www.nytimes.com/... Katrina Jones / @katrina_hrm : “The algorithms are supposed to reduce the burden on understaffed agencies, cut government costs and — ideally — remove human bias. A recent United Nations report warned that governments risked “stumbling zombie-like into a digital-welfare dystopia."" https://www.nytimes.com/... @princeton : The topic of #AIHumanRights shaped a conference at @PrincetonCITP in 2019 as social scientists, activists, policy experts and computer scientists offered their insights to @Alston_UNSR for his @UN report on tech in social protection systems. https://www.princeton.edu/... #AIHumanRights https://twitter.com/... Kevin Baron / @defensebaron : Computers? We left this up to computers? https://www.nytimes.com/... via @NYTimes @bigbrotherwatch : Almost 200,000 people in Bristol are automatically monitored, ‘risk scored’ + flagged by predictive algorithms for likelihood of antisocial behaviour. Our investigations found multiple local authorities using risk score systems - especially in welfare. https://www.nytimes.com/... Helen Gym / @helengymatlarge : In a justice system already biased against black, brown and low-income people, we don't need algorithms making decisions to jail or release people. @mediamobilizing and @mediajustice compiled a nationwide database of these algorithms. https://www.nytimes.com/...
Context & Ripple Effects
This piece lands mid-arc. As far back as 2017, US justices were already weighing what it means that sentencing, bail, and guilt decisions lean on proprietary third-party recidivism scores no one outside the vendor can audit. By late 2019, an MIT-built game using real defendant data was letting players feel how COMPAS misjudges risk firsthand.
What the Times adds is geographic scope: Philadelphia, Rotterdam, and Bristol — where roughly 200,000 residents are automatically monitored and risk-scored — showing this is not a US-only experiment but a transatlantic governing practice, with Big Brother Watch among the groups now demanding human oversight.
First-order effects
- Defendants, probationers, and Bristol residents have patrol routes, sentence inputs, and probation rules set by software they cannot inspect or contest at the moment of decision.
- Civil-society opponents like Big Brother Watch gain a concrete, named target list — police departments, courts, and probation systems in three countries — to press for human review.
Second-order effects
- Defense lawyers and civil attorneys respond on the legal front: the same period produced litigation strategies aimed at automated systems denying basic services, and later saw defense counsel challenge the accuracy of tools like Cybercheck used in thousands of cases.
- Vendors of proprietary scoring tools face a credibility squeeze — the COMPAS critique showed their outputs can be gamed and probed publicly, forcing them toward disclosure or losing procurement battles.
Third-order effects
- If the oversight push holds, criminal-justice AI moves from quietly procured black boxes to audited infrastructure, with transparency requirements becoming a purchase condition rather than an afterthought.
- The deeper structural question the coverage keeps circling: when sentencing and probation inputs are trade secrets, accountability shifts from judges to vendors — a governance gap regulators on both continents will eventually have to close.
The trend: Criminal-justice algorithms are crossing from quiet adoption into contested governance, where litigation, accuracy audits, and human-oversight demands increasingly shape what governments may deploy.