CHINA TOPIX

11/02/2024 03:24:18 am

Make CT Your Homepage

New Software Can Detect Liars 75% More Accurate Than Humans

What does lying look like? Researchers reveal results.

(Photo : University of Michigan) What does lying look like? Researchers reveal results.

A new type of software is being developed by scientists that can detect deceit and lies without having to interact with the subject in question, which is a far cry from traditional types of lie detector devices.

Like Us on Facebook

Researchers from the University of Michigan claim that this one of a kind lie detecting software can sense any sort of dishonesty while it examines the subject's gestures and words. The software based its analysis from 120 video footage from media coverage taken from actual high stakes court trials.

Other videos were also obtained from The Innocence Project which is an organization that aims to absolve those victims who are wrongfully convicted, according to researchers. The footage also contain both testimonies from defendants and witnesses. Half of the footage reveal how the subject appears when lying. 

Researchers tested a prototype of this software and resulted in a 75 percent accuracy in detecting deception, based on the results of the trials. They then compared the software's accuracy to humans' ability to detect deceit which only resulted in just above 50 percent accuracy.

Scientists were also able to identify several behavioral patterns that revealed that a person is lying. Individuals who move their hands more often are apparently lying and these people will also most likely to sound more certain and look directly into the eyes of the questioner more often than not than those who are presumed to tell the truth.

Liars also appear to scowl or grimace more during questioning where this behavior was observed 30 percent of the time based on lying footage of the trials as opposed to 10 percent of the truthful footage.

According to lead researcher, Rada Mihalcea who is also a Computer Science and Engineering professor, it is quite challenging to develop a setting that can motivate people to lie under laboratory conditions when the stakes are not high enough. Mihlace adds that they can reward people who can lie well during experiments and pay them to convince another that something false is true. However, in the real world, deceiving requires true motivation.

During the experiments, researchers analyzed the testimony of the witnesses and defendants from the video footage with the verdicts of the trials to determine which subjects are deceptive.

The researchers also transcribed the audio and counted vocal segregates like "ah", "uh" and "um" and did an assessment how they frequently used specific words. They also counted gestures in the footage with a standard coding scheme for interpersonal reactions where it rates nine different motions for eyes, hands, brow and mouth. 

The next step for this software is to integrate physiological parameters like body temperature fluctuations and respiration and heart rates which will be scanned using non invasive thermal imaging.

Real Time Analytics