AI pattern decodes false pain expressions
In the ever-expanding contest between artificial intelligence and the ordinary human mind, you can chalk up another one for the computer.
Scientists have developed a computer system with sophisticated pattern recognition abilities that performed much better than humans in differentiating between people experiencing genuine pain and people who were just faking it.
In a study published in the journal Current Biology this week, human subjects did no better than chance - about 50 percent - in correctly judging if a person was feigning pain after seeing videos in which some people were and some were not.
The computer was right 85 percent of the time. Why? The researchers say its pattern-recognition abilities successfully spotted distinctive aspects of facial expressions, particularly involving mouth movements, that people generally missed.
"We all know that computers are good at logic processes and they've long out-performed humans on things like playing chess," said Marian Bartlett of the Institute for Neural Computation at the University of California-San Diego, one of the researchers.
"But in perceptual processes, computers lag far behind humans and have a lot of trouble with perceptual processes that humans tend to find easy, including speech recognition and visual recognition. Here's an example of a perceptual process that the computer is able to do better than human observers," Bartlett said in a telephone interview.
For the experiment, 25 volunteers each recorded two videos.
In the first, each of the volunteers immersed an arm in lukewarm water for a minute and were told to try to fool an expert into thinking they were in pain. In the second, the volunteers immersed an arm in a bucket of frigid ice water for a minute, a genuinely painful experience, and were given no instructions on what to do with their facial expressions.
The researchers asked 170 other volunteers to assess which people were in real discomfort and which were faking it.
After they registered a 50 percent accuracy rate, which is no better than a coin flip, the researchers gave the volunteers training in recognizing when someone was faking pain. Even after this, the volunteers managed an accuracy rate of only 55 percent.
The computer's vision system included a video camera that took images of a person's facial expressions and decoded them. The computer had been programmed to recognize that one kind of facial movement combinations suggested true pain and another kind suggested faked pain.
"It's looking at what 20 facial muscles are doing in every frame of video," Bartlett added.
'SADNESS AT A FUNERAL'
So why are people so lousy at spotting a faker? The human face transmits an abundance of information including expressions of emotion and pain. But people also are adept at simulating emotions, some are so good they routinely can deceive others.
"Human facial expressions sometimes convey genuinely felt emotions, and some other times convey emotions not felt but required by a particular social context, for example expressing gratitude after receiving a terrible gift or sadness at a funeral," said the University of Toronto's Kang Lee, who studies lying in children and adults and was one of the scientists who conducted the research.
The computer system proved far better than people at spotting subtle differences between involuntary and voluntary facial movements that underpin sincerity, the researchers said.
"We can envisage in the very near future a widely available and inexpensive computer vision system that is capable of recognizing subtle emotions," Lee said by e-mail.
"Such a system can not only be used to detect deception to prevent medical fraud or to help homeland security but also recognize emotional states of patients who may not be able to communicate very well due to impairments or inability."
Such a system also potentially could be used in law enforcement and screening job applicants, the researchers said.
Bartlett co-founded a San Diego-based start-up company called Emotient Inc to find commercial applications for the facial expression recognition system, focusing on the retail and healthcare fields, she said.