For all the new debate about artificial intelligence, it’s easy to forget that AI technology has been weaving itself into public life for years — and it hasn’t always played out as anyone expected. There’s a type of AI technology, for instance, that cities have spent millions of dollars on and deployed since the 1990s: Gunshot detection systems. For decades, those have been advertised to law enforcement as a way to cut police response times by detecting live gunshots more accurately and frequently than residents can. They use AI algorithms to pinpoint potential gunfire and distinguish it from other loud urban sounds. ShotSpotter, the biggest name in the business, reportedly has a network of at least 25,000 microphone sensors across more than 160 cities, including New York, Boston and Denver. Once the sensors find potential gunfire, human staffers confirm the report and alert 911 call centers or officers’ devices — a process that takes less than a minute, says ShotSpotter’s manufacturer. This month ShotSpotter lost its second-largest client when Chicago Mayor Brandon Johnson, a Democrat, declined to renew the city’s contract. Chicago police’s use of ShotSpotter had been controversial from the start, with watchdogs and public safety groups calling the network unreliable and potentially dangerous to residents. Johnson’s cancellation is the latest blow to a technology long scrutinized for ties to stop-and-frisks, ineffectiveness, high costs, racial bias, dead-end deployments and slower police response times. Chicago joins a small group of cities that have ditched ShotSpotter, including Atlanta and Portland. However, the company boasts a 99 percent renewal rate from its customers over the last four years. Its manufacturer, SoundThinking, says that cities often regret cutting the service. Four cities have reinstated the technology, and all of them even expanded their original coverage area, according to a SoundThinking spokesperson.The spokesperson told DFD that “the handful — literally handful — of customers that let the service lapse often return when they experience the safety implications of life after ShotSpotter.” For some critics, ShotSpotter has become a clear example of the risks of new technologies like AI in the public sector — which sound appealing at first, but can become entrenched despite ongoing doubts about their effectiveness. Politically, it’s easier for leaders to stick with any technology that seems to help safety, said Jonathan Manes, an attorney with the MacArthur Justice Center, which published a study critical of the system in 2021. “These systems end up being sticky not because of their efficacy or usefulness necessarily as a public safety tool, but because people would like to think that there's a technological silver bullet that we can pay for,” said Manes, whose study found ShotSpotter doesn’t work as advertised (the company commissioned a review that disputed its methodology). In 2021, for instance, Syracuse reactivated ShotSpotter at the request of its mayor Ben Walsh, an independent, just eight months after disbanding the program due to pandemic budget cuts. Walsh’s challenger in the upcoming mayoral race took credit for the switch, saying he “made it an issue” by raising a ruckus over the loss of ShotSpotter before a spate of shootings. Even in Chicago, there’s a political twist to the decision: Johnson made canceling ShotSpotter a campaign promise, to win progressives’ votes. But he chose Sept. 22 as the official end date — factoring in that the city would host the Democratic National Convention in August. The system will still be active when the convention comes to town. It’s also hard for cities to determine, reliably, if a system like ShotSpotter works. SoundThinking says the technology works well; based on self-reported data from all ShotSpotter clients, it touts a 97 percent accuracy rate and 0.5 percent false positive rate — figures it commissioned a third-party data analytics firm to verify. It also trots out success stories in which alerts have helped officers find gunshot victims. Its defenders in law enforcement have praised ShotSpotter alerts in terms of the number of firearms recovered, pieces of evidence collected, and lives saved. But for a city considering whether to buy or cancel the system, there is no agreed-upon metric for what positive outcomes should gauge ShotSpotter’s effectiveness — be it lives saved, shorter response times, more arrests or fewer firearm homicides. That makes it difficult for a city to evaluate its use of ShotSpotter between contracts, or swap it for technology from newer vendors, said Ed Vogel, a member of national and Chicago-based #StopShotSpotter campaigns. “There’s a structural problem at play here,” he told DFD. “We're seeking tools to solve a problem, but the problem has not been named.” For cities that end their ShotSpotter contracts, its presence is entrenched another way: physically. Chicago first installed ShotSpotter in 2012 over a three-square-mile area that grew, through two renewals and tens of millions of city dollars, to over 100 square miles of coverage by 2018. The system of sensors “becomes a durable infrastructure in a city,” Vogel said. According to past ShotSpotter contracts, it’s up to the company whether to remove these components if there isn’t a renewal, or just leave them in place. Governments that reinstate the service later just won’t have access to reviewers’ alerts from the lapsed period. Vogel said that within those terms, it’s possible for the company to leave its mics up, continue recording audio “as a way to train its algorithm,” and convince the city to later reintroduce an upgraded version — a strategy that reportedly unfolded in Fall River, Mass. “They left them out for free as a way to try to entice the city to renew with this idea that they're going to improve the accuracy,” said Vogel. “There's probably going to be something similar happening in Chicago.” The SoundThinking spokesperson did not deny the possibility, but said each case is unique, and the company evaluates “the best way to move forward” once a contract has expired. Johnson’s office did not respond when asked if the sensors will be removed when Chicago stops using the technology.
|