The University of Arizona Alumnus / Winter 2009
deception detection
by Tim Vanderpool
Jacob Chinn photos
At first glance, Kevin Moffitt sure seems like a straight shooter. But first glances don’t count for much down here, in the buzzing bowels of the Eller College of Management.
Today, the doctoral student is sitting in a plastic chair, as banks of cameras fix him in their cold gaze and a legion of video monitors reveal his slightest flush or tiniest twitch. A slender laser beam zeroes in on his carotid artery, waiting for a nervous spike of his pulse; another lens traces the dilation of his iris, looking, always looking, for a man about to deceive.
In this well-wired lair, the slightest minutia of human response is sliced, diced, dissembled, and dissected, all with the stunningly complex goal of simply distinguishing truth from lie.
“We’re trying to bring some science to what people do in a poker game,” says Jay Nunamaker, management information systems (MIS) wunderkind and Regents’ Professor in the Eller College. “Everybody thinks they can analyze what people are doing, that they know who’s bluffing. But the literature from thousands of studies show that — when faced with a single scenario — people are only 54 percent accurate in (spotting) deception.” By contrast, Nunamaker says, his equipment “has consistently gotten results in the 80 percent range.”
But Nunamaker’s project is not actually about poker. It is only the most tactile aspect of a broad project that places The University of Arizona on the forefront of America’s national security efforts: the UA was recently designated as co-leader of a new Center of Excellence for Border Security and Immigration. This prestigious appointment by the U.S. Department of Homeland Security comes with $16.5 million in shared funding over six years, and the promise of making the university a major player in this growing field for years to come.
The UA attracted the right staff and drew on key expertise in departments across campus for the new center. Its work ranges from cutting-edge identification optics, to the probing analysis of immigration policy, to balancing streamlined cross-border trade with the demands of heavy security. In the future, you might see the fruits of this work in airports, as you’re subtly screened by a chatty, kiosk-screen avatar while retrieving your ticket. You might witness newly efficient border ports, where commercial trucks breeze through with cargos that have been preregistered, sealed, and electronically monitored.
The UA’s contribution might result in a fuller, more nuanced understanding of immigration’s impacts on our country, or the ability to better predict a terrorist strike.
Here in Nunamaker’s basement, however, those lofty efforts are boiled down to the basics of detecting human skullduggery. The goal is an ability to screen vast numbers of people in a short time — such as in an airport or a border port — without laying a hand on them. At the same time, this work aims to augment often fallible human judgment with the precision of computer analysis.
It’s a high-stakes contest between electronic wizardry and the timeless art of deception, where victory comes in consistently and correctly parsing terrorists or criminals from the masses of traveling humans. That demands sifting through staggering amounts of data at breakneck speeds, and then bundling the information into a form that’s immediately useful to screeners. Still, for all the bells and whistles, everything boils down to simply sniffing out the truth.
Consider Kevin Moffitt. He’s simply sitting in that chair, his face monitored by the phalanx of hardware, from the multiple cameras to microphones that catch his every peep. Another doctoral student, Chris Diller, is adjusting one of the monitors.
“We’re tracking three video cameras on one person, with the point of evaluating them on a level playing field,” Diller says. “All of them use different techniques for deception detection.”
Eventually, this technology will allow screening to occur at various degrees of intensity, but without the physical contact required of, say, wiring someone to a polygraph machine.
“That just wouldn’t be practical when screening large groups of people,” says Nunamaker. “We’re trying to capture all of these things, from a distance, noninvasively, using more linguistics and motion and kinesis. We’re really on the leading edge of that work.”
Still, these remarkable efforts in security technology raise an equal number of privacy and civil liberties questions. Already, the Department of Homeland Security (DHS) has come under fire for confiscating and scanning laptop computers at ports-of-entry, and for advocating measures such as “enhanced” drivers’ licenses that would contain vast amounts of personal information.
While recognizing those concerns, Nunamaker says such considerations lie far beyond his job description: “The DHS has hundreds of lawyers looking at this — at whether you can do these things or not.” And he suggests, subtly, that the delicate balancing act between national security and personal privacy will never completely be laid to rest. At the same time, he notes that this emerging technology might actually enhance privacy, since it would be less physically invasive than many current airport screening procedures, and could drastically reduce the incidence of racial profiling as other “cues” become paramount.
Elyse Golob is the center’s executive director, and she sits in a hushed conference room at the UA’s Science and Technology Park, across town from Nunamaker’s subterranean lab. She says the university laid plans to compete for the DHS designation early on.
“There were rumors that this was coming. So we were well-prepared and well-positioned when it did. We’d taken an inventory of border-related research going on in the various colleges here, and there was certainly a wealth of expertise, from the Department of MIS in the Eller School, with Jay Nunamaker, to Hsinchun Chen in the Artificial Intelligence Lab.” Chen is known for his work on Dark Web, a federally funded project that searches the Internet for terrorist-affiliated Web sites, and compiles them into a vast database.
This tour de force likewise reaches into the College of Optical Sciences and its research on sensors and fiber networks, and the College of Engineering’s development of vehicle identification systems. It draws from the College of Social and Behavioral Sciences, the Center for Latin American Studies, and the Udall Center for Studies in Public Policy, where researcher Judith Gans analyzes the economic impacts of immigration.
Meanwhile, in a field so vital, intense, and time-sensitive as national security, keeping up with ever-shifting priorities presents its own challenges, prompting what Golob calls a process of constant negotiation.
“The agency has problems they want solved, from (surveillance) towers to the optimal number of agents to be deployed on the border. They want answers to those problems yesterday, as well they should. But the university’s basic role is to answer the next generation of issues that are going to occur along the border, working with DHS to anticipate what challenges are in the pipeline.”
This pipeline can get down to nuts-and-bolts issues, such as creating communication networks along the border — between agencies and within the agencies themselves — that actually work the way they’re supposed to. For example, “How do you make a sensor that’s placed on top of a cactus talk to a sensor that’s on a Border Patrol jeep?” Golob asks. “How do you make a communication system that’s coherent?”
There’s also the rather grim work of delineating risk levels — a task that’s become profoundly more complex in recent years. “For instance, the Twin Towers were inconceivable in 2000,” Golob says. “Now they’re planned for. What are those risks we don’t even imagine right now that might happen? We have to develop a model that’s predictive of that risk.”
On a brighter note, there’s another element lurking behind all this hard work: remarkable economic opportunities for the UA. The seeds for that were planted long ago, says Bruce Wright, the university’s associate vice president for economic development. In fact, they reach back to the early 1990s, when the North American Free Trade Agreement was being hammered out between Canada, the United States, and Mexico.
At that time, “a group of people came to us and asked to assess how Arizona could play a meaningful role in the emerging agreement,” Wright says. The result was the Arizona-Mexico Commission, “which was created to look at trade and commerce.
“It also became clear that technology could play a huge part in addressing this issue,” he says. That initiative led to a National Institute of Justice grant, used to study technologies at the border, and gave birth to the concept of a Border Security Technology Commercialization Center in Southern Arizona.
“So we’ve been dealing with this issue for a long time,” says Wright, “and we’ve built up considerable expertise in this area.”
The program aims to enhance prosperity along the U.S.-Mexico border, a place where more than one million people cross back and forth each day, and serious socio-economic and environmental issues are endemic. Improving that situation, while simultaneously enhancing security, requires a delicate balancing act.
That’s quickly obvious back in the basement of the Eller School. By now, Kevin Moffitt is squirming slightly in his seat; a screen, nearby, shows his face draped in various colors. A screeching speaker behind a curtain is meant to startle him, as another screen uses infrared to track his gaze; if he lingers on an object such as a partially constructed bomb long enough, it is enough to tip off the monitors.
In another lab, not far away, doctoral student Mark Patton is constructing a kiosk, of a type that may become common in airports of the future, where passengers are screened by the avatar. Detecting even the smallest nuance that could raise a flag is key.
“We eventually hope to do another system that will include real-time analysis,” Patton says, “fusing the different data streams in real time, to give some type of a probability — some type of deception index — on a scale of one to 20. And the higher the number, the more likely the person is lying.”
And therein lies the point. No matter how polished the prevaricator, there are always clues. Nunamaker calls it the “leakage theory.”
“The key we’re operating under is that you’ve got hundreds and hundreds of cues,” he says. “Linguistic, vocality, psychometric cues — heart rate, the face heating up. And cues are good predictors.
“But while humans can only focus on one or two cues at a time, agents at the border are now trying to track hundreds at a time.”
There must be a better way. And that’s the point of this work. But security research from a business college?
“Oh, it’s still MIS,” Nunamaker says with a slight grin, “although I guess you’d call it MIS hardcore.”
Back to Winter 2009 contents page