Detecting lies has never been a foolproof science. Current methods can provide an educated guess but none are considered conclusive.
Now, new research from the University at Buffalo has found a promising new way to identify untruths.
UB researchers have developed a computer program that measures eye behavior - using cameras that shoot 30 frames per second.
The eyes have it
Human interrogators can identify a lie accurately about 65 percent of the time, according to Ifeoma Nwogu, an assistant professor at UB. These lie-detecting professionals look for changes in behavior, trying to read facial cues for anything that can inform their guess.
But there are no hard and fast rules, says Nuwogu.
“You have rumors, like, when people lie they look you straight in the eye or they’re being shifty-eyed,” Nwogu says. “But there’s no one specific cue that generalizes across liars.”
A team of researchers at UB’s Center for Unified Biometrics and Sensors (CUBS) has set out to tackle that problem.
The CUBS researchers decided to film faces - tracking every little facial movement - to try to identify tell-tale cues. They devised an experiment where test subjects lied and told the truth in the course of a single conversation.
When breaking down the footage, Nwogu says patterns in facial behavior became clear.
“It appears that a lot of the involuntary motion we make when we try to conceal our emotions happens in the upper part of the face,” she says.
Researchers isolated footage of the eyes and were able to establish a baseline of normal truth-telling behavior with each test subject.
When researchers later observed footage containing falsehoods, they designed an automated computer program that could quantify changes in behavior.
Telling lies creates a cognitive conflict, Nwogu says, which is very difficult to prevent from expressing itself in some physical manner. Yet, a small number of test subjects were able to maintain consistent behavior throughout the experiment.
“In over 80 percent of those who lied [82.5 percent], there was a sharp change in the way their eyes behaved. But it was very momentary. It was something a human could have missed,” Ngowu says.
“But because we were using at a computer looking at 30 frames per second, we very clearly saw there was a marked difference when people lied.”
A promising start
This method is not yet considered definitive or accurate enough to go mainstream, Nwogu admits. Only 40 people comprised the first research test group, so more experiments with larger groups will give more insight into its reliability.
Even then, isolating eye behavior will likely complement, not replace, existing truth-testing methods.
Plus, with most research focusing on the eyes for now, there’s still the matter of studying the rest of the face.