… I don’t buy it. It’s horrifically easy to mess up this kind of experiment, say by taking all of one class’ photos at a different time or day, or with a different device. They claim validation; but I also expect way more natural error (we miss mild Autism diagnosis, no?).