In recent years, facial recognition has become one of the most widespread and powerful technological tools. For those with newer models of Apple iPhones, simply unlocking the phone requires facial recognition. Many individuals incorporate facial recognition into their daily lives in routine ways, such as by having Google or Apple Photos create albums of different people in their camera roll. However, personal use is not the focus of backlash against this developing surveillance tool. What people are worried about is how large corporations and law enforcement will collect and use the resources provided by facial recognition.
In Eugene and Springfield, Flock Safety cameras— which scan and track license plates— were banned after Eugene Police Department identified unspecified vulnerabilities and limitations regarding the AI powered technology. Retailers, such as Macy’s, Amazon Go, Kroger, Target and Lowe’s, are some of the large corporations that have experimented with facial recognition software. Rite Aid, for example, attempted to use facial recognition to identify potential shoplifters but was met with a ban from the Federal Trade Commission (FTC) after it falsely accused women and people of color. Since facial recognition software has historically been trained on white males, women and people of color are more likely to be misidentified. This vulnerability, coupled with the tendency for law enforcement to view black males as dangerous, could lead to even more false accusations for people of color. Facial recognition is also easily trained and adapted to inherit certain biases, so if law enforcement doesn’t fairly regulate their technology, it could lead to increased racial profiling.
Additionally, these misidentifications are not uncommon, as many cases have been reported, like that of Nijeer Parks. When law enforcement in Woodridge, Chicago, sent a low-pixelated driver’s license photo into facial recognition software, the computer misidentified the suspect, leading to the arrest of Nijeer Parks, who had to spend ten days in jail. Facial recognition wasn’t the only tool law enforcement could have used in that case. There was other evidence available at the shoplifting crime scene, such as fingerprint marks and DNA from a discarded water bottle, but authorities pushed away these leads and instead depended on a computer to uphold the law. This is dangerous, as reliance on machines could lead to unflagged systematic discrimination. Furthermore, if law enforcement has already messed up with small crimes like shoplifting, imagine the consequences if the computer misidentified an individual in a large scale investigation.
If the identification is the problem, one might just suggest that the computers need more training. For more training, the software will need more photos. Where organizations will obtain more photos is a point of contention. If grocery stores, Ring cameras (which directly partner with law enforcement through their Neighbor app) and other unavoidable businesses are using facial recognition technology, then there is virtually no effective way to mandate consent, and it would be a violation of privacy. It is tedious and unreliable to require each person to sign off on facial recognition consent when they just want to grab a few groceries and get home. Currently, many law enforcement databases use mugshots or driver’s licenses to identify suspects, but mugshots are taken before individuals are convicted of crimes, so they have not explicitly given consent for their photo to be used in these investigations.
In conclusion, the vulnerability for facial recognition software to misidentify suspects, especially people of color, and the privacy risks of training and using the software should be considered in the face of such a technologically-evolving world.
