A team of researchers at the University of Michigan (UM) said it is working to develop unique software that can detect lie by considering a speaker’s gestures and words.

The researchers said they are building the software based on real-world data and for that they studying the videos from high-stakes court cases. The software, so far, has shown 75% of accuracy in identifying who was being deceptive as compared with humans’ scores of just above 50%, said study researchers.

They also noted that the individuals who lied moved their hands more frequently. They even tried to sound more certain. The individuals who lie also looked their questioners in the eye a bit more often than those presumed to be telling the truth, among other behaviors.

In order to develop the software, the team is used machine-learning techniques and trained it on a set of 120 video clips from media coverage of actual trials.

Rada Mihalcea, a professor at UM who leads the project with Mihai Burzo, assistant professor at UM-Flint, said, “In laboratory experiments, it's difficult to create setting that motivates people to truly lie. We can offer reward if people can lie well, pay them to convince another person that something false is true. But in real world there is true motivation to deceive”.

In the clips of people lying, the researchers found some common behaviors, such as scowling or grimacing of the whole face. These behaviors were seen in 30% of lying videos versus 10% of truthful ones, said researchers.