Computer outdoes humans for spotting fake faces

Kathleen Blanchard's picture
Could you spot a fake facial expression better than a computer could?
Advertisement

Recognizing facial expressions is a sign of emotional intelligence and important for understanding how others are feeling. We rely on people's faces to tell us if they're in pain, telling the truth, happy, sad and more. Researchers from the University of Toronto have discovered computers are better than people at detecting false facial expressions. What do the findings mean?

Facial expressions can be deceiving

According to the researchers, people can deceive us with their facial expressions. The scientists say computers could be better at detecting pain malingering.

The finding that is published in the journal Current Biology means it could become harder to fake pain.

Marian Bartlett, research professor at UC San Diego’s Institute for Neural Computation and lead author of the study said in a press release: “Human observers just aren't very good at telling real from faked expressions of pain.”

Advertisement

Machine vision systems could reveal "fakers”. “Bartlett said "...our approach has the potential to elucidate ‘behavioral fingerprints’ of the neural-control systems involved in emotional signalling."

For instance, a person that is faking a facial expression opens their mouth frequently, but without much variation. The researchers say watching a person's mouth is the most predictive of a false facial expression.

“Further investigations,” said the researchers, “will explore whether over-regularity is a general feature of fake expressions.”

Bartlett said spotting false facial expressions with computer software could also have implications for homeland security, law and more, including spotting psychopaths. Facial expressions could also provide important clues about health, attentiveness, sleepiness and more that people try to hide.

Even with training, the researchers found computers can detect false facial expressions better than humans. The computer detected fake emotions with 85 percent accuracy compared to trained humans whose accuracy was only 55 percent. You can test your own ability to spot a fake expression by taking the Berkeley Emotional Intelligence quiz.

Abstract:
Current Biology
March 20, 2014

Advertisement