The software is up to 75% accurate in detecting a liar.
Researchers from the University of Michigan have developed new software that can detect a person’s lies.
Don't Miss: Sam's Club Black Friday 2016 Details
Unlike the polygraph, this lie-detecting software does not need a person’s pulse or breathing rate to catch a lie instead it considers the speaker’s words and gestures to determine whether they are lying or not.
Researchers examined videos of high-stakes court cases to create the software, though; it is still in a prototype stage right now.
During experiments, it was up to 75% accurate in identifying deceptiveness when it was checked with trial outcomes, whereas humans were found just 50% accurate in detection.
So how does this software actually works? Researchers say they studied a set of 120 video clips of actual trials and identified some common behaviors in those who are lying. For instance, lying individuals moved their hands more, tried to sound more certain, spoke vocal fills such as ‘um’ quite frequently and looked directly in the eyes of the questioner more often than a person who is telling truth.
The key here to develop the software is ‘real-world data.’
"In laboratory experiments, it's difficult to create a setting that motivates people to truly lie. The stakes are not high enough," said Rada Mihalcea, one of the authors of the study. "We can offer a reward if people can lie well – pay them to convince another person that something false is true. But in the real world there is true motivation to deceive."
The trial videos include testimony of both defendants and witnesses. It is an obvious thing that both sides cannot be right at the same time, one of them must be telling a lie.
To find out which one is, researchers transcribe the videos including vocal fillers and analyzed how often subjects used various words. Then, they looked at the gestures of the subjects and used a standard coding scheme to interpret different motions of eyes, head, brow, mouth and hand.
Researchers fed the data into their system and found it was 75% accurate in identifying who is lying. The score was much better than the guesses of humans.
“People are poor lie detectors,” said Mihalcea. “This isn’t the kind of task we’re naturally good at. There are clues that humans give naturally when are being deceptive, but we’re not playing close enough attention to pick them up. We’re not counting how many times a person says 'I' or looks up. We're focusing on a higher level of communication."
Researchers are aiming to incorporate physiological parameters such as heart rate, respiration rate and body temperature fluctuation into it to draw more accurate conclusions. They are currently training the computer to classify various gestures.