There have been numerous cases, all over the world, but most prominently in the United States, and most usually in casinos, where cops have wrongly arrested and charged completely innocent people based on a complete misconception that many cops (and sometimes other authorities) have.
And that misconception is that machine facial recognition is extremely reliable and almost foolproof. Certainly more reliable than people's subjective estimations.
Indeed, a lot of casinos in the United States (and sometimes some other establishments) have facial recognition software running that tries to detect people who have been banned from said casinos for one reason or another. When they ban someone, they take pictures, feed them to the software, and the software then uses facial recognition algorithms on all the live feed that they get from their hundreds of security cameras, and alerts the staff if there's a positive, ie. someone who has been banned potentially being back.
The thing is, a lot of that security camera footage is surprisingly low-resolution and blurry, filming the premises primarily from ceilings and overall from quite a distance. The reliability of the facial recognition software is very poor, particularly when the pictures are blurry.
Yet, there have been numerous cases were casinos have called the cops based only and solely on the false positive of such a facial recognition software, and the cops actually went ahead and arrested and charged the person, who was actually completely innocent of any wrongdoing (ie. he wasn't actually banned from the casino and it was just a case of misidentification by the facial recognition software.)
(And if you are wondering "why don't they check his ID to see if he's the same person", they always suspect that the person is using a fake ID, so it isn't helpful to them, at least not in that situation.)
The curious thing in these situations is that the cops relied completely on the accuracy of the facial recognition technology even though they themselves could see with their own eyes that the person in the camera footage was not the person they were arresting. In some of the bodycam videos released about these cases the cops quite explicitly state that they are arresting the guy because the machine had recognized him as the banned person. And that's it. That's all the evidence they needed for the arrest. In fact, I once saw such a video where the cops were actually doubting out loud whether it was actually the same guy, because he looked so different from the security camera footage, yet still proceeded to arrest and charge him anyway
Those cops quite clearly showed a rather curious psychological phenomenon where they think that machines can't be wrong. That eg. facial recognition made by a machine is much more accurate and reliable than humans visually comparing faces. That the machine is impartial, objective, accurate, engaging in mathematical precision well beyond human subjective capabilities, and thus is extremely reliable and can't be wrong.
Somehow those cops have never been educated nor informed about the utter unreliability of facial recognition software, and how (especially back when those arrests were made) they are significantly worse than human perception and capability of recognizing faces. They had somehow formed the notion that "machines can't be wrong", that they only engage in mathematical precision and thus are pretty much perfect, and certainly a hundred times more reliable and accurate than humans. You can clearly see this mentality at play in those bodycam videos.
Comments
Post a Comment