Researchers from the University of Michigan (U-M) are using machine-learning techniques to develop new lie-detecting software.
The team is training the software on video of media coverage of actual court trials. The prototype considers the words and gestures of the speaker, and does not need to touch the subject in order to work.
The group reports the software was up to 75% accurate in identifying who was being deceptive, as defined by trial outcomes, compared with only more than 50%.
The software associated more hand movement, trying to sound more certain, looking questioners in the eye, and other behaviors with lying individuals. "There are clues that humans give naturally when they are being deceptive, but we're not paying close enough attention to pick them up," says U-M professor Rada Mihalcea. She says the software could be a helpful tool for security agents, juries, and mental health professionals.
The initiative is part of a larger project to integrate "physiological parameters such as heart rate, respiration rate, and body temperature fluctuations, all gathered with non-invasive thermal imaging," says University of Michigan-Flint professor Mihai Burzo.
From The University Record
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA