Can AI read Micro Expressions?

"Micro Expressions" have been a hot topic for a long time, featured in the TV show "Lie to Me" and so on. The idea goes back at least to Charles Darwin, who hypothesized that some facial expressions are involuntary. When presented with something disgusting enough, Darwin thought, our faces would register disgust no matter how much we wanted not to. Once that first reactive moment is past we might be able to erase the telltale face and impose a different one, but because the first reaction is automatic it could not be completely eliminated.

Current science seems to indicate that this is true, and people have been trying to build lie detecting systems based on these responses. On the other hand it is also true that lying often causes a galvanic skin response, and yet that sort of lie detector has all sorts of problems. Critics think systems based on facial tells will do no better. After all people vary a great deal in how readily their emotions show on their faces. Good poker players are probably much better at hiding their reactions than the rest of us, and of course there are hysterical people who react much more strongly than average.

The EU is actually testing an AI system for detecting lying faces in airports right now, but so far as i can tell they have not released any results. I would be willing to bet that these systems will not work very well for identifying terrorists or whatever they are worried about, and will instead deliver tons of false positives from all the people who find airport security a terrifying ordeal.