Lie-detecting Kiosk Can Catch Deceptive Travelers
A lie-detecting artificial intelligence program developed at SDSU is being tested at airports and borders around the world.
“Most of us are terrible at detecting deception, even experts,” said San Diego State University management information systems professor Aaron Elkins. “We often are overconfident and rarely are proven otherwise, so we just assume that we’re really good at detecting deception.”
That overconfidence can spell big trouble when lives depend on determining who’s being deceitful, such as at international security checkpoints at borders and in airports. To overcome humans’ poor lie-detecting abilities, Elkins is working on a robotic lie detector that can accurately sniff out deception.
The lie-detecting kiosk is known as an Automated Virtual Agent for Truth Assessments in Real Time (AVATAR).
“AVATAR is a kiosk, much like an airport check-in or grocery store self-checkout kiosk,” Elkins said. “However, this kiosk has a face on the screen that asks questions of travelers and can detect changes in physiology and behavior during the interview. The system can detect changes in the eyes, voice, gestures and posture to determine potential risk. It can even tell when you’re curling your toes."
Though it’s not being used in any official capacity yet, the device is being tested at several international locations including Canada and Singapore.
Here’s how it works: Passengers would step up to the kiosk and answer a series of questions such as, “Do you have fruits or vegetables in your luggage?” or “Are you carrying any weapons with you?” Eye-detection software and motion and pressure sensors would monitor the passengers as they answer the questions, looking for tell-tale physiological signs of lying or discomfort. The kiosk would also ask a series of innocuous questions to establish baseline measurements so people are just nervous about flying, for example, wouldn’t be unduly singled out.
Once the kiosk detected dishonesty, they would flag those passengers for further scrutiny from human agents.
Last year, Elkins and his team performed pilot experiments in an airport in Ottawa, Canada. For the study, 82 volunteers participated in a mock customs interview, half of whom were assigned to try to carry through a package of “drugs”—though there weren’t really any drugs involved, of course.
The participants stepped up to the AVATAR kiosk and were asked a series of 17 questions about why they were traveling and what they were carrying. Each corner of the screen displayed icons representing guns, other weapons, explosives or drugs, respectively. The kiosk tracked their eye movement, vocal modulation and body movements, looking for the subtle signs that the traveler was uncomfortable answering the question.
In this particular case, eye-tracking proved the most effective method for finding the liars, noticing that the would-be smugglers had a particular pattern of avoiding looking at the drug icon on the screen. But Elkins stresses that each scenario has different contexts that make some lie-detection methods more effective than others.
Different cultures also have different responses to questioning, he noted, so those have to be accounted for as well.
“The overarching premise behind all this is there is no Pinocchio’s nose,” he said. “You need to have these multimodal measurements, combined and fused.”
Elkins is continuing to partner with government agencies to test out the kiosk’s abilities in real-world situations.
“AVATAR has been tested in labs, in airports and at border crossing stations,” Elkins noted. “The system is fully ready for implementation to help stem the flow of contraband, thwart fleeing criminals, and detect potential terrorists and many other applications in the effort to secure international borders.”
