Last Wednesday, June 10, the tech giant Amazon announced that it was banning police agencies from using its face recognition technology Rekognition for a year. The company gave no official reason for its timing, although it comes less than two weeks after the death of George Floyd at the hands of the Minneapolis police force.
Critics have charged that face recognition technology in general, and Rekognition in particular, has biases when it comes to race or gender. According to an Associated Press report, in the past Amazon has defended its software against the results of studies by MIT researchers who showed that such software marketed by Microsoft, Amazon, and IBM sometimes made mistakes that put darker-skinned people and women at a disadvantage. (Microsoft and IBM pledged to address the deficiencies as a result.) Amazon’s product is a fairly minor player in a field that includes products from Japan and Europe as well, which many police agencies use. Both Amazon and other producers have called for federal regulation of face recognition technology amid concerns of abuse and bias.
Another portion of the human anatomy has been used for identification by police and security agencies for decades, largely without recent controversy: the unique patterns on fingertips. But at least so far, fingerprints cannot be sensed remotely, and taking them usually requires the cooperation or at least physical contact with the individual being examined. With the proper optical technology, one can image a face from a distance of a kilometer or more, leading to the possibility of mass identification in crowds with software that can unambiguously connect faces to persons.
Speaking from an engineering ethics viewpoint, we can identify the parties in this controversy as follows. There are the firms that make face-recognition technology. There are the customers and potential customers for such technology, which include but are not limited to law-enforcement agencies, governments in general, and also private firms wanting to make personalized advertisement appeals, for example. There are regulatory agencies with the potential ability to regulate such technology. And then there is the general public, a subset of whom are criminals, but the vast majority of whom are just ordinary people trying to live their lives in these lately rather extraordinary times.
The dangers of misusing face-recognition technology are many, even if it works perfectly. In China, for example, it is already being used in combination with other techniques to monitor movements and activities of the general public in what we in the U. S. would consider gross violations of privacy. But even if the agency using the technology was entirely benign (and if there are real human beings running it, it won’t be entirely benign), flaws in the technology such as a preferential tendency to make false positive identifications of darker-skinned people will lead to injustices such as innocent persons being identified, and possibly arrested and worse, in connection with wrongs they did not commit.
So concerns like these are probably behind the motivation that made Amazon ban its Rekognition software from police use for a year. Such concerns comprise at least one reason that face-recognition firms have called for federal regulation of the technology as well.
However, it does seem rather perverse for a company to defend a product they made for police departments, only to turn around and take it away from them for a year. And here I wish to tread lightly, because in today’s “cancel culture,” anything you say can be used against you, as the police used to say (and maybe still say, for all I know). Only you used to have to be formally charged with a crime for that to be the case, but no longer. Anyway, here goes.
The publicity surrounding George Floyd’s death has led to both intense outrage expressed in print, in words, and in massive demonstrations, as well as to criminal acts such as looting and rioting. Understandably, a lot of the hostility inspired by Floyd’s death has been directed against law enforcement agencies ranging from local police organizations up to and including the federal government. Recently, some have called for “defunding” police forces, the interpretation of which varies, but at a minimum means a punishing cut in financial support.
Order is a fundamental need for any functioning society. It is not the ultimate good of society, which lies elsewhere, but it is needed as much as food and water. Because there are always a few people who will not follow the rules simply because someone in authority tells them to, most functioning societies of any size have law-enforcement agencies.
The modern concept of law includes the principle that it applies equally to everybody. Sometimes the police fall short in trying to achieve that ideal, and certain groups that include minorities receive unfair treatment. Every reasonable means should be used to try to remedy such wrongs, and to get closer to the ideal of equal treatment under the law.
But to penalize entire organizations for the wrong acts of a few of their members is to erode the very thing we are trying to achieve, namely, equal treatment under the law. A crippled (or, perish the thought, abolished) police force will lead to increased disorder, and a reaction that may well institute a much harsher order than any of us want.
Speaking specifically of Amazon and Rekognition, if the software really didn’t work that well, Amazon should never have sold it to the police in the first place. Taking it away for a year looks suspiciously like the time I took away my 10-year-old nephew’s toys for a day to punish him for not minding his aunt and uncle. And one can certainly question the right of a private company to punish police departments, which are under the authority of those governmental divisions that control them, not Amazon.
Sources: The Associated Press article about Amazon’s banning Rekognition from police use appeared in numerous news outlets, including the Tampa Bay Times on June 12. The argument about the rule of law was inspired by an interview I heard with Princeton professor Robert P. George on the Sheila Liaugminas Relevant Radio network show “A Closer Look.”
Republished from Engineering Ethics.