Facial Recognition Technology: The Need for Robust Civil Rights Protections
By Livia Luan
The past few years have witnessed the dramatic expansion of the United States’ facial recognition technology market, which is growing by 20% each year and is expected to be worth $9 billion a year by 2022. A type of biometric technology that analyzes images of human faces in order to identify them, facial recognition technology can benefit society by verifying the identity of government employees at high-security facilities, combating child sex trafficking, finding missing persons, and more.
It also has great potential for harm, as evidenced by recent studies that demonstrate higher error rates for people of color and for women. In an MIT Media Labs study, computer systems using facial images to recognize skin color and gender could correctly classify lighter-skinned men 99% of the time, but could only correctly classify darker-skinned women 65% of the time. In addition, the ACLU conducted a test in which Amazon’s facial recognition service incorrectly matched 28 members of Congress with mugshots — disproportionately misclassifying members who are people of color. Higher error rates occur when benchmark data used to train facial recognition algorithms are skewed towards a subset of the population, overrepresenting lighter-skinned men while underrepresenting darker-skinned people in general.
The pervasiveness of algorithmic bias across numerous tech companies’ facial recognition systems suggests that the way we look could be weaponized against us, especially in light of the rise of sophisticated surveillance networks throughout the country. In Detroit, a $1 million real-time facial recognition system allows police officers to identify and track residents captured on hundreds of private and public high-definition cameras at locations such as schools, churches, abortion clinics, and addiction treatment centers. In Chicago, the police department and transit authority use a system that compares faces captured on video surveillance footage to the city’s database of around seven million mugshots. Moreover, federal law enforcement agencies have forged alliances with officials at departments of motor vehicles in states like Utah, where agents with the Federal Bureau of Investigation and Immigration and Customs Enforcement logged more than 1,000 facial recognition searches between 2015 and 2017.
Given the existence of higher error rates for people of color and for women, the deployment of facial recognition technology as a surveillance and public safety tool must be scrutinized for its potential harms. For example, law enforcement officials who rely on flawed data to initiate immediate-response and street-level interactions could end up subverting due process protections, arresting innocent individuals, and damaging community trust. Furthermore, facial recognition surveillance could deter individuals, particularly members of historically-surveilled immigrant communities and communities of color, from exercising their freedoms of speech, assembly, and religion. During protests in 2015, for example, the Baltimore police may have used facial recognition software and a private company’s analysis of social media accounts to identify and arrest people with outstanding warrants.
Not all cities have welcomed the use of facial recognition technology for surveillance purposes. San Francisco and Somerville, MA voted in May and June, respectively, to ban government use of facial recognition software. At the federal level, Congress is in the process of examining the technology’s implications through hearings and legislation. As lawmakers at multiple levels of government explore ways to regulate the technology, support for its use is declining among registered U.S. voters. According to a Morning Consult/Politico poll conducted June 14–16, 2019 among 1,992 registered voters, 42% said they generally support its use, reflecting a decrease of seven percentage points from the 49% who supported its use in an August 2018 Morning Consult survey.
In seeking a method to regulate the use of facial recognition technology, we must recalibrate the balance of power between government and the people in a way that protects constitutional rights while preserving the promise of technology. To do so, we need to view this issue through gender and racial equity lenses, in recognition that the harms of this technology disproportionately threaten women and people of color. Therefore, legislation should be judged for the ability to guarantee strong civil rights protections and to promote transparency, accountability, and equity.
Until comprehensive criteria for regulating facial recognition technology are fleshed out, it should not be used for surveillance purposes. In June, Asian Americans Advancing Justice | AAJC — as well as over 60 privacy, civil liberties, civil rights, and investor and faith groups — signed onto a letter urging the House Committee on Oversight and Reform to “put in place a federal moratorium on face recognition for law enforcement and immigration enforcement purposes until Congress fully debates what, if any, uses should be permitted.” We understand that this technology disproportionately affects communities of color, and are deeply worried that within the Asian American and Pacific Islander community, historically-monitored groups such as South Asian Americans and Muslim Americans are most at risk for scrutiny. Therefore, we stand with other advocates in pushing Congress to address our collective concerns before facial recognition surveillance becomes too prolific and difficult to regulate.
Livia Luan is the programs associate and executive assistant at Asian Americans Advancing Justice | AAJC, where she supports the telecommunications, technology, and media program on rapidly evolving issues such as digital privacy, digital equity, and facial recognition technology.
