What cities must do to responsibly deploy facial recognition solutions
To help cities as they grapple with the pros and cons of facial recognition programs, the National League of Cities has published recommendations for guiding conversations around the emerging technology and how it is implemented.
While not every city that uses facial recognition has voted on a policy, cities are taking the lead in shaping how the technology can best be implemented, NLC said in its April 21 report.
First and foremost, NLC urges cities be transparent about their intentions. They should insist on community input on any decisions and require public officials to vote on use of the technology before it is rolled out to law enforcement. Policies should be available online, and citizens should be able to easily submit complaints or questions about the government’s use of facial recognition.
Second, law enforcement should be thoroughly trained not just on the technology, but also on its potential for bias. The use of high-quality photos is encouraged as is independent reviews of system-generated matches. Officers should not be able to make an arrest based solely on facial recognition, NLS cautions, and they should understand the consequences of misusing the technology.
As part of its third recommendation, NLC recommends use of facial recognition be limited to serious or violent crimes to reduce the risk of misidentifications and privacy violations. To avoid violating the privacy of innocent citizens, cities should consider installing a system that alerts police only when surveillance cameras capture a suspect’s face. Cities should also carefully weigh the pros and cons of driver’s license or mug shot databases as sources for photos.
Fourth, it is critical cities have rigorous standards to protect biometric databases from privacy or cybersecurity violations. Photos or videos not relevant to investigations should be deleted, as should mug shots of those found to be innocent. Data should not be stored indefinitely. Access to biometric databases should be restricted, and both two-factor authentication and forensic tracking of system use should be enabled.
Finally, contracts with facial recognition vendors should insist companies certify their algorithms are based on demographically representative training sets and that they regularly test their algorithms for accuracy and bias. By carefully reviewing indemnification clauses, cities can be sure the vendor will be held responsible for any errors. A pilot program is recommended before cities sign any long-term contracts with vendors.
“The conversation concerning facial recognition is particularly sensitive given the technology’s imperfections and how it is frequently implemented and used behind closed doors,” the report states. “By following these recommendations, cities can better facilitate public discussions about facial recognition technology in their communities.”
More information on what facial recognition is, how cities are using and regulating it and how city officials can best approach public conversations about facial recognition is available in the full report.
Connect with the GCN staff on Twitter @GCNtech.