How facial software can identify liars
Facial-recognition software could warn immigration officers if someone is lying
Surveillance cameras at airports could be used to identify passengers who are not telling the truth, helping better identify terrorists and drug-smugglers while minimising instances of racial profiling.
The artificial-intelligence programme was developed by a team of computer scientists at the University of Rochester, New York, using crowdsourcing technology to build the largest public data resource of liars’ facial expressions.
Researchers used a machine-learning algorithm to analyse over 1.3 million frames of one-to-one interactions interactions.
Among the most common traits associated with lying identified by the programme was a high-intensity version of the so-called Duchenne smile, in which people effectively smile with their eyes. Another surprising finding was that honest people would contract their eyes, but not smile with their mouths.
Even highly-trained immigration officers often struggled to assess whether someone is lying, deliberately concealing something or just nervous.
However, existing lie-detector technology is “controversial and a camera-based system that could be more widely used would be even more so”, says The Times, “especially if it found its way into courtrooms, offices and schools”.