Fatal Self-Driving Crash, Wrongful Arrest Expose Smart City Tech Risks

Two machines built to make city life safer did the opposite. One, a self-driving Uber, struck and killed a woman walking her bicycle across a Tempe, Arizona street in March 2018, the first death from a collision involving an autonomous vehicle. The other, a facial recognition system, misread a surveillance image and helped police wrongfully arrest a Black man in Detroit for a crime he did not commit. Both cases sharpen a question cities keep struggling to answer before new technology lands on their streets: will it genuinely help the people who live there, or mainly serve the companies selling it?

A new paper published in the journal Cities takes a hard look at that tension. Researchers from Arizona State University, the University of Massachusetts Amherst, and Chalmers University of Technology in Sweden examined the real-world case studies, self-driving cars in the Phoenix metro area and facial recognition software in Portland, Oregon, to show how city governments have struggled to balance what tech companies want to sell against what residents actually need.

Their argument is simple. City halls across the country have faced pressure to adopt flashy “smart city” tools, from traffic sensors to surveillance cameras, often pitched with promises of efficiency and safety. But when local governments lack a clear way to weigh those promises against community values like safety, equity, and privacy, the gap can turn into real harm. To help close that gap, the researchers adapted an existing public administration framework called public value mapping, originally developed in 2007 by scholar Barry Bozeman to evaluate science and public policy, and tested it against the messy, real histories of self-driving cars and facial recognition.

How Two Cities Tested Smart City Technology

To make their case, the authors did not run a lab experiment or survey thousands of people. Instead, they took a close look at two specific situations: the rollout of self-driving cars in the Phoenix region, with a focus on Tempe, and the rise and eventual banning of facial recognition technology in Portland.

Their research drew on city websites, mission statements, laws, policies, planning documents, city council records, news coverage, and conversations with city staff in both locations. From these sources, the team identified what each city said it valued, things like public safety, equity, or transparency, and then tracked a series of “policy events,” such as an executive order or a company’s product launch, to see whether each one pushed the city closer to its stated values or further away from them. Every event was also weighed against how much it benefited the private companies involved, since a technology can succeed commercially while failing the public, or the other way around.

Phoenix Opened Its Roads to Self-Driving Cars

Arizona’s path started with an open door. In August 2015, Governor Doug Ducey signed an executive order allowing self-driving car testing on the state’s public roads, drawing in companies thanks to what the paper describes as a “hands off” regulatory approach. Google began testing in Chandler, and in February 2017, Uber started testing its own self-driving vehicles in Tempe. Local leaders largely welcomed the technology as a possible fix for road safety and traffic problems.

That optimism did not last. On March 18, 2018, an Uber self-driving car struck and killed Elaine Herzberg as she walked her bicycle across a Tempe street, the first fatality from a collision involving an autonomous vehicle. A federal investigation eventually spread blame across Uber, the vehicle’s safety driver, and the Arizona Department of Transportation. Governor Ducey suspended Uber’s testing in the state and updated his executive order to require companies to meet existing safety standards, but according to the paper, no new safety standards were actually created in response to the crash.

Even so, the region did not walk away from the technology. Tempe formed a committee to study how self-driving cars might support the city’s broader goals around safety and quality of life. Nearby, Chandler partnered directly with Waymo, letting city employees use its ride-hailing service for official business and training local police and fire departments on how to interact with the vehicles. Waymo also teamed up with the regional transit agency, Valley Metro, to test using self-driving cars to fill gaps in public transportation. Researchers point to these partnerships as examples of a city and a tech company finding ways to create benefit on both sides, even after a tragedy exposed how much was still unknown about the technology’s real risks.

Portland Chose to Ban Facial Recognition Technology

Portland took a different road entirely. Facial recognition technology, which identifies people by scanning their face in photos or video, had already drawn criticism for working unevenly across racial groups. A 2019 study by the National Institute of Standards and Technology found that some facial recognition systems were up to 100 times more likely to misidentify Black and Asian faces than white faces.

Those numbers turned into a real-world case in January 2020, when a Black man in Detroit was wrongfully arrested after facial recognition software linked him to video footage from a 2018 shoplifting incident. A later review found loose standards and a lack of safeguards in the identification process. Detroit’s police department responded by updating its policies to forbid using the technology for surveillance and predictive purposes.

Portland had already built a smart city office, Smart City PDX, focused on equity and community input, and it leaned on that foundation to respond to growing concern from residents, especially communities of color, about the risks of facial recognition. After a public engagement process, the Portland City Council voted unanimously in September 2020 to ban the use of facial recognition technology both by the city itself and by private businesses in public spaces, going further than cities like Oakland and San Francisco, which had only banned municipal use. Portland kept building on that policy, adopting a broader surveillance technology policy in 2023 that requires privacy reviews before new surveillance tools can be used or purchased.

A Tool to Score Smart City Technology Before It Causes Harm

Out of these two contrasting stories, the researchers make the case for wider use of public value mapping. Rather than a simple pass-or-fail test, it lays out, side by side, whether a piece of technology or a policy decision mainly helps businesses, mainly helps the public, helps both, or helps neither. The goal isn’t to hand city officials an automatic answer, but to give them a structured way to talk through tradeoffs with residents, companies, and elected leaders before, during, and after a technology rolls out.

Applied to Tempe and Phoenix, the mapping shows a region that started by prioritizing economic growth with little attention to safety, suffered a serious failure on both fronts when a pedestrian died, then slowly moved toward partnerships that better balanced public and private interests. Applied to Portland, it shows a city that identified its values early through community engagement and then acted decisively once facial recognition was shown to conflict with those values.

Researchers are upfront about the limits of this approach. Public values can be hard to pin down and may shift over time, and simply naming a value does not guarantee it will be honored in practice. Still, they argue the tool gives city staff, who often lack the time, money, or expertise to evaluate the flood of new technology being marketed to them, a workable way to ask the right questions.

What happened in Tempe and Detroit were not abstract policy debates. They were a death and a wrongful arrest, the kind of outcome that can follow when governments embrace powerful technology faster than they can weigh its risks. Portland’s experience shows a different outcome is possible when a city builds in that check from the start, listens to the communities most likely to be harmed, and is willing to say no to a technology even when it’s marketed as inevitable. Cities adopting the next wave of smart technology, whatever form it takes, would do well to ask who actually benefits before residents end up paying the price.