- San Francisco’s ban on the use of facial recognition technology by municipal agencies is noteworthy given the city’s high-tech affiliation and AI’s potential applications in public safety.
- The safety-enhancing benefits of facial recognition are not resonating; instead, the technology has become a lightning rod for societal concerns related to privacy and inequality.
San Francisco is set to become the first major U.S. city to ban the use the facial recognition technology by municipal agencies. On Tuesday, the San Francisco Board of Supervisors voted in favor of the ‘Stop Secret Surveillance Ordinance,’ outlawing the use of the AI-based technology by city departments. The move is particularly noteworthy because it originates in a part of the U.S. otherwise known for embracing high tech and because it restricts the use of artificial intelligence for public safety, widely considered a top use case for facial recognition technology. However, San Francisco isn’t the only city evaluating restrictions on facial recognition; the issue is top of mind among lawmakers in many regions.
Opponents of the use of facial recognition cite concerns over individual privacy. It’s no surprise that individuals are concerned about the amount of information that is being compiled about them. Greater public awareness of the details being collected by applications such as Facebook, and how that information is being disseminated, has alarmed many individuals and contributed to distrust of new information-gathering technologies. Nonetheless, intensifying efforts by various organizations to collect personal information seems to be inevitable and increasingly difficult to avoid. The ability to track cell phones via GPS remains a contentious issue. Today, mobile device users can activate location-based services to improve application performance, but Android and Apple have made privacy settings available for users that do not want their location tracked or shared, and will share location data only for calls to emergency services numbers. With facial recognition, the discussion will likely need to center on acceptable use (by governmental organizations and enterprises) and the ability to opt-in (for enterprise applications), and it will undoubtedly include use limitations established by law.
Opponents are also concerned about recent reports that facial recognition models are biased or prone to inaccuracies. Some studies suggest that facial recognition’s performance is inconsistent and subject to greater error when analyzing women and darker-skinned individuals. All AI models are only as good as the data used to train them, and much of the training of facial recognition models has been done using images of white males. To apply the models to a broader community, they will need to be trained with a more diverse data set. This can be done; it just takes time and needs to be prioritized. However, given the sensitive nature of how facial recognition could potentially be used (and misused) by governmental organizations, it makes sense to wait until models improve before the technology is widely deployed.
Interestingly, the safety-enhancing benefits of facial recognition are not resonating well with the general public. Perhaps the technology is a victim of poor timing. Facial recognition is coming to market during a period when the public is disgusted with the way social media has been using and sharing their data. Similarly, there has been a public outcry against the unequal treatment of women in the workplace and of minorities by law enforcement. As a result, the technology appears to be a lightning rod for contempt, exemplifying the biases and inequalities still prevalent in our society.
Society’s attitudes towards facial recognition may change over time as the technology evolves. We are still in the very early stages of its adoption, and as with all emerging technologies, there will be important issues that must be addressed before moving ahead. For the time being, many communities are likely to err on the side of caution, restricting the technology’s use until its performance is more consistent and society has set norms regarding acceptable use.