Unlike a polygraph test, it doesn’t need to touch a subject to work.
Using real-world data from high-stake court cases, researchers from the University of Michigan are developing an impressive way to catch someone in a lie. While polygraphs have to be hooked up to a subject to work, their new software doesn’t even need to touch the subject in order to pinpoint a lie. In the experiments with the prototype, the software detected which individuals were lying with a 75 percent accuracy rate — for reference, humans can only successfully detect lies about 50 percent of the time.
Instead of measuring a person’s pulse and breathing rate like in a polygraph, the software focuses more on a person’s behavior by considering words and gestures. The researchers say they’ve identified “several tells” of lying throughout their research. For example, liars tend to move their hands more, and they try to sound more certain of themselves. Interestingly, in the video data from the court cases, the lying subjects looked their questioners in the eye more than those who were presumed to be telling the truth.
To develop the lie-detecting software, the researchers trained it with machine-learning techniques on a set of 120 video clips from media coverage of real court trials. Some of the clips came from The Innocence Project, a national organization that works to exonerate those who have been wrongfully convicted of a crime.
The team also transcribed the audio of the videos, including vocal fills like “um, ah, and uh.” Additionally, they counted the subjects’ gestures using a standard coding scheme that scored nine different motions of the head, eyes, brow, mouth, and hands. Since the research includes “real world” data from actual court cases, it is more compelling than other studies on lying.
"In laboratory experiments, it's difficult to create a setting that motivates people to truly lie. The stakes are not high enough," Rada Mihalcea, professor of computer science and engineering, said in a press release. "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."
After feeding the data into the new system and leaving it to sort out the videos, it was 75 percent accurate at identifying the liars based on their words and gestures. While some of us might like to think we have more than a 50/50 chance at pinpointing a liar, most humans simply don’t possess this skill.
"People are poor lie detectors. This isn't the kind of task we're naturally good at,” said Mihalcea. “There are clues that humans give naturally when they are being deceptive, but we're not paying 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."
Although we may not be lie-detecting naturals, the researchers highlighted 6 key behaviors to look out for in liars:
Scowling or grimacing of the entire face — in 30 percent of lying videos vs. 10 percent of truthful ones
Looking directly at the questioner — in 70 percent of lying clips vs. 60 percent of truthful
Making gestures with both hands — in 40 percent of lying videos vs. 25 percent of truthful
Speaking with vocal fills like “um, ah, and uh” was more common among liars
Using words like “he” or “she” rather than “I” or “we” in order to distance themselves from the wrong action
Using phrases that reflected certainty was more common among liars
All of this effort is part of a larger project. The researchers are working on integrating the physiological aspects of lying into the program as well, like heart rate, respiration rate, and body temperature fluctuations. They’re gathering this data with non-invasive thermal imaging.
The researchers are also taking cultural influences into account. "Deception detection is a very difficult problem," Mihai Burzo, assistant professor of mechanical engineering at UM-Flint, said in the press release. "We are getting at it from several different angles."
The researchers hope the system will be useful for security agents, juries, and even mental health professionals. Unfortunately for those of you hoping to use it on a suspected cheating partner, you’ll have to use your 50/50 human lie-detecting abilities.