AVATAR Kiosk Designed to Help Law Enforcement at Borders

The AVATAR kiosk uses non-invasive artificial intelligence and sensor technologies developed by the Center for the Management of Information at the Eller College.
Feb. 16, 2011
The AVATAR kiosk uses non-invasive artificial
intelligence and sensor technologies developed by the Center for the Management of Information at the Eller College.
The AVATAR kiosk uses non-invasive artificial intelligence and sensor technologies developed by the Center for the Management of Information at the Eller College.

Travelers were abuzz about new airport technology over the holiday season, but a different approach is under investigation at the University of Arizona as part of the Department of Homeland Security-funded, 14-university BORDERS consortium.

"The problem of security is not going away," said Doug Derrick, a fifth-year doctoral student in the University of Arizona's management information systems department. "It's clear that we need some form of automation to accommodate increasing border traffic."

Enter the AVATAR kiosk.

The kiosk is an interactive screening technology designed to be on the front lines of border crossings and airports. Individuals would approach the AVATAR kiosk, scan their identification, answer a few simple questions and then move on.

Meanwhile, the AVATAR kiosk has used non-invasive artificial intelligence and sensor technologies to gauge suspicious behavior. Law enforcement would step in to work directly with individuals whose behavior has been flagged.

"The goal is not to replace the person, but to augment law enforcement's ability to detect deception," said Derrick.

In its current, initial iteration, the AVATAR kiosk assesses cues through sensors in body movement, vocalics, pupillometry and eye tracking.

"We have 400-500 psychophysiological and behavioral cues that we can assess for detection," said Jay Nunamaker, Regents and Soldwedel Professor of MIS, Computer Science, and Communication. "We're continually looking for a subset that's a reliable and accurate predictor."

The AVATAR kiosk is currently looking for 15 different cues out of 500 that are a good predictor for detecting deception. Some of the cues – such as eye tracking – are completely involuntary.

"Humans can maybe control two or three cues at one time, but not all 15," Nunamaker said. "So it's inevitable that tell-tale cues will leak out no matter how hard people try to control them when being interrogated."

Nunamaker and Derrick put the kiosk to the test earlier in the fall when they traveled to Warsaw, Poland, with a prototype and ran an experiment with a participating group of European Union border guards.

Half of the guards constructed mock bombs and packed them in suitcases, then tried to get through the AVATAR kiosk. AVATAR detected all of the mock bombers, but captured some false positives as well.

"We set the threshold pretty high," Derrick explained. "We wanted to catch all of the ‘guilty' parties. But clearly we have more work to do in reducing those false positives."

The AVATAR kiosk is the culmination of a collective 42 years of research in the areas of deception detection and collaborative work conducted by Nunamaker and professor Judee Burgoon, director of human communication at the Center for the Management of Information.

Nunamaker is principal investigator and director of the BORDERS consortium, which is headed by the UA. Derrick, who entered the doctoral program in MIS at the UA after working for a defense contractor, is developing his dissertation around the animated avatar that interacts with users through the kiosk.

"The current version of the avatar is a first cut; it's not a perfect design," he said. "We want to see how people respond to avatars of different ages, ethnicities, voices, hair colors, demeanors. My dissertation used gender and demeanor as a starting point."

Derrick tested user response to a smiling female face, a smiling male face, and neutral versions of each: "We measured testers' reactions to the avatars, looking at areas including trustworthiness, likability, and power/dominance."

In general, the neutral male avatar was viewed as more powerful, and the female avatar was viewed as more likable, but ultimately the avatar may have dissimilar affects on individuals being screened.

"This finding has important implications for future screening practices," Derrick said. "For example, human agents may select different avatars based on the individual being screened. Cultural considerations are also significant and must be taken into account."

"The idea is that down the line, the avatar will be able to detect human response and adjust accordingly," said Nunamaker.

When more information is included on the kiosk, such as a fingerprint reader and a magnetic strip reader, Derrick said,

"The avatar will have access to information that will allow for a more robust interview; for example, if someone scans an Italian passport, the avatar will address the person in Italian."

"We will be moving from a fixed script to a dynamic script," Nunamaker said. "But it's important to note that this technology is not ready to go out into the field. There is so much lab work and field work yet to be done – we're years away. Right now the sensors we're using are in a large room, and it's a huge engineering feat to pack them into this kiosk."

AVATAR is just one project under investigation as part of the BORDERS consortium project. The project as a whole was nominated for a Governor's Award for Innovation this year, and the AVATAR was featured on the Discovery Channel program Daily Planet.