A look at New York City's slow attempts to increase transparency around its use of algorithms, which govern many policy areas, including policing
Rebecca Heilweil / Vox :
Context & Ripple Effects
New York City had already spent years wiring algorithms into everyday governance before this piece ran: the Drive Smart pilot tracked the driving habits of 400 volunteers back in 2015, and by 2023 the MTA was running AI-powered surveillance at seven subway stations to catch fare evaders. The Vox report lands in that gap between deployment and disclosure — algorithms shape policing and other policy areas, but residents largely cannot see which systems are in use or how they were vetted.
The slow pace matters because the accountability tools arrived only piecemeal afterward: the city's [[a:840498|law requiring firms using AI in hiring to notify candidates and submit to annual bias audits]] covers private employers, not the city's own systems, while a 2018 child-welfare scoring tool is still drawing racial-bias concerns years into operation.
First-order effects
- City agencies running opaque systems — most sensitively in policing — face renewed pressure from council members and advocates to disclose what algorithms they use, on what data, and with what oversight.
- Residents subject to algorithmic decisions, from policing to benefits screening, still have no general right to know when a model shaped the outcome affecting them.
Second-order effects
- Vendors selling algorithmic systems to the city will increasingly be asked to contract for auditability and bias testing, as the hiring-law model of independent annual audits becomes the template buyers reach for.
- Advocacy groups gain a concrete benchmark: every new deployment like the MTA's fare-evasion surveillance now gets measured against the transparency standard the city keeps failing to meet for its own tools.
Third-order effects
- If the pattern holds, US cities converge on a two-track regime — statutory notification-and-audit rules for private-sector AI, while government's own systems remain governed by slower, discretionary review — leaving public-sector algorithms the least scrutinized category.
- Municipal algorithm registries and disclosure laws pioneered in New York become the reference point other cities copy, making vendor audit practices a de facto national standard even without federal regulation.
The trend: Cities are adopting algorithmic decision-making faster than they build the transparency machinery to govern it, with accountability arriving system-by-system rather than by design.