|Credit: Image courtesy of University of California - San Diego|
I’ll reveal the answer a little later. But for now, would it surprise you to know that most people are no better than chance at distinguishing actual pain from simulated pain? Even after training, human observers were only 55% accurate in picking out who was only faking. Computers, on the other hand, have no such limitations and could distinguish genuine from faked expressions 85% of the time.
Scientists led by Marian Bartlett of the University of California, San Diego and Kang Lee of the University of Toronto compared human observers with a computer vision system called CERT (Computer Expression Recognition Toolbox). CERT detects human faces and codes the positions and actions of facial muscles using a program called FACS (Facial Action Coding System). FACS scores incremental movements in individual facial muscles.
The models for the experiment were asked to place their hands in either ice water (pain) or warm water (fake) for one minute. Their facial expressions were recorded and played back for either the human observers or the CERT program. As mentioned, the humans were pretty much guessing whereas the computer was actually quite good at spotting the fakers.
I’m not sure how CERT’s pain-detecting abilities would be used in the real world. I’m envisioning mothers saying: “So, you can’t go to school because you have a stomach ache? We’ll see about that.”
Now for the answer. If, like me, you guessed that image B on the right was fake, you’re wrong. Image B shows the expression of real pain. It’s image A that’s fake.