By Caroline Bottger
Facial recognition technology at an ethical inflection point: should it be banned outright, or regulated? In an attempt to find a middle ground, law professors from New York University and the University of the District of Columbia proposed five regulations that would permit its use, with significant guardrails. We’ll take a look at these regulations below and see where they could be improved.
Starting out, the op-ed makes the distinction between “face surveillance” and “face identification,” “Face surveillance” is the real-time tracking of people’s movements as they go about their day, regardless of whether a crime has been committed; “face identification” is the active identification of a criminal during an investigation. The former should be banned, but the authors lay out a case for regulating the latter.
The first requirement is that face identification should only be used until all races and genders can be identified correctly. This is sensible: it’s well past due for face recognition to improve its accuracy, with incidences of misidentification being well-documented in the press. But this comes with another security stumbling block: According to the New York Times, many people have no idea that their faces are stored in the databases used to train these algorithms, raising further privacy issues that will need to be addressed sooner rather than later.
Secondly, face identification should only be used for the most serious of crimes, such as murder, rape, robbery and assault. This would limit the use of facial recognition since violent crime in the United States is at record lows. But it shouldn’t be ignored that facial recognition makes law enforcement’s everyday job much easier, especially for lower-level infractions like credit card fraud and road rage incidents. Taking away technology that streamlines the more mundane cases might be a hard sell.
Thirdly, face identification should not be limited to criminal databases. The authors wrote that “mugshot databases are the product of decades of discriminatory policing for offenses like drug crimes,” and using other databases would lead to better recognition accuracy. This train has left the station, though. Private databases have already sprung up to complement government ones, and are used in both face surveillance and face identification scenarios. But providing more transparency into these private databases could build trust with the public.
Fourth, face identification should only be employed with a judicial warrant. Law enforcement officers already obtain warrants for searching property, so bringing face identification into the fold makes sense. Because as easy as it is to identify faces using facial recognition, there should be safeguards for proper use. “Like police wiretaps, such [database] searches have a higher risk of being unjustly invasive,” wrote the Salt Lake Tribune Editorial Board. But since people move around, and houses don’t, these warrants would have to be granted with some speed.
Lastly, any law permitting face identification should include penalties for misuse. This is reasonable, and would go a long way in building trust with the public. Standards for misuse would have to be set, too.