Researchers hope that the AI system called Deception Analysis and Reasoning Engine (DARE) could be used in courtrooms to discover whether or not people are telling the truth.
The machine was developed by researchers from the University of Maryland and Dartmouth college who trained the system using videos of people in court.
In their study, published on the Cornell University Library website, the researchers, led by Dr Zhe Wu, said: ‘On the vision side, our system uses classifiers trained on low level video features which predict human micro-expressions.”
The team trained the system to recognise micro-expressions associated with lying, these include: frowning, eyebrows raising, lip corners turning up, lips protruded and head side turn.
After watching 15 courtroom videos, the AI system was able to work out if someone was lying.
Results showed that DARE could accurately spot 92 per cent of the micro expressions.
The same test was then given to humans, of which only 81 per cent recognised tell-tale signs of fibbing.
Researchers concluded that DARE was better at detecting if someone was lying than humans.
They said: “Our vision system, which uses both high-level and low level visual features, is significantly better at predicting deception compared to humans.”
The results also suggested that the DARE could be even more effective if the system was provided with more information.
Researchers said: “When complementary information from audio and transcripts is provided, deception prediction can be further improved.”
VISIT THE SOURCE ARTICLE
Author: || World Economic Forum