San Diego State University assistant professor Aaron C. Elkins has developed a new type of airport kiosk that can help authorities find travelers who are lying, smuggling, or otherwise engaged in possibly criminal activity. The Automated Virtual Agent for Truth Assessments in Real Time (AVATAR) utilizes technology that, like its human counterparts, observes and assesses unconscious physical and vocal cues that indicate deception.
“AVATAR is a kiosk, much like an airport check-in or grocery store self-checkout kiosk,” Elkins told SDSU's News Center. "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."
The Washington Post notes that the kiosks analyze data gathered through changes in voice, pupil and eye movements, facial expression, and posture.
Travelers coming to security checkpoints, as they might in clearing customs or arriving at a border, stand before the AVATAR unit and answer a series of questions, starting with those that establish a baseline by asking for simple personal information. Once that baseline is established, other questions that might elicit discomfort or physical signs of deception from a liar come into play. These signs of deception are nearly impossible to mask. "It’s a lot harder to lie because you’re using strategies,” Elkins told The Post. “You’re managing your story, you’re managing your impression, you’re evaluating the perception of yourself in a kind of meta way, and then you’re changing your story if you think it’s successful or people think you’re suspicious." Those who are flagged for possibly deceptive answers are further interviewed by human officials.
In an interview with Seeker.com, Elkins declined to discuss what types of questions AVATAR might ask, but a look at Elkins' patent for the technology AVATAR is based on and some of his published work on vocalics – the analysis of voice stress during deception – allows for a glimpse at how the technology works.
According to the patent, human test subjects were asked 4 "blocks" of 4 questions each by avatars of varying genders (male/female) and demeanors (neutral/smiling) after being "instructed to pack a bag of various items on a table to take through a screening checkpoint." Although it is unclear whether or not these questions are similar or identical to those currently in use, it's easy to see how some them could elicit a stress response. The first block of four questions was given as follows:
"Please describe in detail the contents of your backpack or purse.
I am detecting deception in your responses. Please explain why that is.
What will you do after you get through this checkpoint?
Please tell me how you have spent the last two hours before coming to this checkpoint."
If this seems excessive, it's because it's intended to be. In a chapter entitled "Unobtrusive Deception Detection" from The Handbook of Affective Computing, Elkins writes that "The ECA [Embodied Conversational Agent] within the AVATAR can also take on any demeanor (e.g., friendly, stern) or embodiment (e.g., male, female) that will elicit the most diagnostic behavior. During the interview, the AVATAR conducts an interview protocol designed to exaggerate the differences in predicted behavior between liars and truth tellers."
And it works. Elkins reports that in experiments, AVATAR was 94% accurate.
While AVATAR has recently gone through testing Canadian Borders Services Agency after earlier tests (for data gathering only) at border points in Mexico and Romania, it's in no danger of replacing humans at the border... at least not yet.
"Computers are really good at measuring small, minute changes in behavior or voice or pupil size," Elkins said in an interview with The Guardian.
"But people are really good at plausibility, and language, and understanding the logic of your story. So the entire system is meant to be kind of a partnership between a human and the computer, leveraging the AI’s strengths in analyzing behavioral or physiological changes that humans simply can’t detect."
Although it's unclear what hurdles AVATAR might face before we start seeing it in American airports, Elkins says that it's ready to ship. “In terms of the technology, I would say that it’s ready,” he told The Guardian. “It’s just a matter of finding the right application and fit.”