How facial software can identify liars

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”.