Facial recognition is not exactly an up-and-coming technology; it’s here and it’s established. Currently, the likenesses of over 117 million American adults are documented in law enforcement facial recognition databases without their consent (Garvie et. al). These images have been used to successfully identify suspects in cases including the 2018 Capital Gazette shooting (Bogost). While lauded by proponents for its potential to create a safer society, there are several unaddressed shortcomings of this technology. Unregulated use of facial recognition technology by law enforcement could lead to the suppression of free speech and erosion of other civil liberties. The recognition technology is far from 100% accurate and misidentifies women and people of color at disproportionately high levels (Lohr). While the United States is in the early stages of this technology use, China is a chilling example of the power that surveillance and perceived surveillance have to control citizens’ behavior. Facial recognition is an important tool for law enforcement, and completely discontinuing its use is unrealistic and likely impossible. However, to preserve individual freedoms and civil liberties, it’s vital that regulations are imposed which prevent overreach and abuse.
The first electronic facial recognition system was developed in the 1960s by Woodrow Wilson Bledsoe. The rudimentary technology required a user to manually input coordinates of a subject’s facial features, creating a template that would be compared and matched with images entered into a computer (History). Since then, facial recognition technology has developed the capacity to identify key features of a face and map them automatically, then compare that map to millions of faces collected from public and private databases (Facial). This evolution represents a broadening of both the scope and the abilities of this technology. According to Georgetown Law’s comprehensive facial recognition investigative project The Perpetual Line-Up, law enforcement facial recognition affects over 117 million Americans. 16 states allow the FBI to use driver’s license and mugshot databases in their searches involving facial recognition. Facial recognition may also be used by local police in field identification, upon arrest, while investigating a crime, or in conjunction with live video surveillance feeds. This evidence is admissible in court and has been used to convict people (Gullo).
In China, the use of facial recognition by law enforcement is prevalent and unregulated. Specific, accurate information about the scope of government facial recognition usage in China is undisclosed, in part because the illusion of widespread surveillance is often enough to manipulate citizens to regulate their own behavior out of fear. However, it is known that there are at least 170 million surveillance cameras operating in China (Jacobs). Facial recognition is used by the government in public transportation, law enforcement, and, disturbingly, the tracking of Uighur Muslims (Byler). In contrast to the high level of importance that Americans place upon personal freedoms and civil liberties, a history of authoritarianism in China influences many Chinese citizens to accept surveillance and government control more readily in the name of security (Li). Opposing the government is dangerous, which is compounded by the lack of anonymity created by facial recognition.
In the United States, facial recognition is used by law enforcement in four main ways. It may be used in a stop, in which someone who is unwilling or unable to identify themselves is photographed and processed through the system which provides a match; in an arrest, in which a mugshot is enrolled in the system; in an investigation, in which a photo of a suspect is obtained and searched against the facial recognition database, producing a list of possible matches; and in live video surveillance, which continuously compares the faces of those passing a security camera running a live feed to a list of people who are being searched for (Garvie et. al). Facial recognition usage is commonly used at airports, sporting events, border crossings, and in the private sector (Garvie et. al).
The Perpetual Line-Up ranks the risk potential of these activities based on their transparency, the scope of people subject to investigation, real-time or “after the fact” deployment, and whether it is an established or “novel” usage. Risk is defined as the potential for violations of privacy, civil rights, and civil liberties (Garvie et. al). Searches that subjects are unaware of, that are continuous and generalized, or unprecedented compared to the use of fingerprinting or DNA analysis are considered especially risky.
Facial recognition technology is a fantastic asset to criminal investigations but contains a critical flaw. It allows law enforcement to identify suspects with up to 99% accuracy — as long as the suspect is a white male. Today’s facial recognition algorithms were trained almost exclusively using images of white men, and as a result, dark-skinned men and women are vulnerable to higher rates of misidentification (Lohr). According to an interview by The Perpetual Line-Up, two unnamed major facial recognition companies do not test for racial bias. In her study, MIT researcher Joy Buolamwini found that gender was misidentified in up to 1% of white men, up to 7% of white women, up to 12% of dark skinned men, and up to 35% of dark skinned women. This disparity is especially concerning when considering that “Blacks are more likely than others to be arrested in almost every city for almost every type of crime”(Heath), and are therefore overrepresented in American mugshot databases.
Chinese persecution of Uighurs, an ethnic minority that primarily practices Islam, represents a dangerous abuse of facial recognition targeting an entire minority group. The Chinese government has marked Uighur people as terrorists, limiting their access to public spaces and services through facial recognition checkpoints which all users are subject to (Byler). Their explicit goal is to suppress and control the Uighur population in the name of national security. Facial recognition is used to restrict their autonomy and freedom of speech, with some systems programmed to track Uighurs exclusively (Authorities). This racial profiling, combined with tracking of their internet use and Uighur internment in re-education camps, shows how facial recognition is a powerful tool for oppression and cruelty in the hands of a government that wishes to control its population.
Residents of nations using facial recognition must confront the conflict it creates between personal privacy and public safety. Facial recognition can create safer societies by ensuring that it’s difficult for criminals to hide from authorities, even overseas. However, the notion that the government has the right to know the location of every citizen at all times—that citizens do not have the right to go out in public anonymously—is one for serious consideration. China’s use of this technology to control Uighur Muslims is an example of how governments may abuse their power. Additionally, many ways in which facial recognition is used are not analogous to ways in which law enforcement historically has used identifying information such as fingerprinting, mug shots, or DNA testing. Given the unprecedented nature of facial recognition and its high value for law enforcement, there is potential for abuse. It’s necessary that the ways in which facial recognition is used by law enforcement are publicly disclosed and regulated.
Though American and Chinese authorities are unlikely to relinquish facial recognition technology, citizens of the United States should harness the power of democracy to demand that legislation is passed which creates ethical standards and enforces transparency for law enforcement’s use of facial recognition. The Perpetual Line-Up proposes 30 recommendations for law enforcement, the legislature, and other entities, several of which are highlighted here.
Recommendations for legislation to be passed mainly aim to broadly define when it is and is not acceptable to use facial recognition. For example, it is proposed that law enforcement only conduct facial recognition searches of driver’s licenses and mug shots on an individual basis, with a court order and probable cause. Facial recognition used with live surveillance feeds should only occur when there is an immediate, “life-threatening public emergency”(Garvie et. al) and a court order. Tracking people because of their race, ethnicity, religious or political views should be prohibited. These suggestions, while not exhaustive, create basic protections that are not currently in place.
Recommendations for law enforcement aim to create transparency about FBI and local law enforcement use of facial recognition. In regards to racial bias, the FBI should regularly test its system for accuracy and bias, and publicly release results. It should also “Investigate state and local agencies’ use of face recognition for potential disparate impact”(Garvie et. al), recognizing and working to prevent disproportionate impact on specific groups. Finally, the FBI and Department of Justice should reverse their proposal to exempt their facial recognition system from Privacy Act requirements that “guarantee Americans the right to review and correct non-investigatory information held by law enforcement—and the right to sue if their privacy rights are violated”(Garvie et. al). Facial recognition has great potential to be used to violate privacy and civil liberties. Therefore laws regulating facial recognition should focus on the public’s knowledge of their enrollment in these systems, how they function, and their usage.
In many ways, facial recognition AI is a double edged sword. Powerful technology can help law enforcement apprehend violent criminals quickly, but can be abused to track and restrict citizens’ autonomy. A large subject pool offers a greater likelihood of finding the suspect, but may lack the consent of its constituents and increase the possibility of false matches between people with similar appearances. Surveillance may aim to create a safer society, but in effect forces the government to regard all citizens as future criminals. Americans should have the right to understand how, why, and when they are being surveilled, especially because the use of facial recognition systems is funded by taxpayers’ money. With degrading privacy standards in China, it’s essential that the United States set the example of a nation that utilizes facial recognition to protect its citizens, not to control them.