Lie Detector Firm Lobbies CIA, DOD on Automated Eye-Scanning Tech

A Utah-based outfit overseen by a former CIA consultant has spent hundreds of thousands of dollars lobbying intelligence and defense agencies, including the CIA and DHS, to adopt its automated lie detection technology, public lobbying disclosures reviewed by The Intercept show. Converus, Inc., boasts on its website that its technology has already been used for job screenings at American law enforcement agencies, corporate compliance and loss prevention in Latin America, and document verification in Ukraine. The company’s management team includes chief scientist John Kircher, a former consultant for the CIA and Department of Defense; Todd Mickelson, former director of product management at Ancestry.com; and Russ Warner, former CEO of the content moderation firm ContentWatch.

Warner told The Intercept that lobbying efforts have focused on changing federal regulations to allow the use of technologies other than the polygraph for lie detection. “The Department of Defense National Center of Credibility Assessment (NCCA) is in charge of oversight of validation and pilot projects throughout the U.S. government of new deception detection technologies,” Warner wrote in an email. “DoD Directive 5210.91 and ODNI Security Agent Directive 2 currently prohibit the use of any credibility assessment solution other than polygraph. For this reason, we have contacted government agencies to consider the use of EyeDetect and other new technologies.”

After finding success in corporate applications and sheriff’s offices, Converus has set its sights on large federal agencies that could apply its EyeDetect technology to a host of uses, including employee clearance screenings and border security. Unlike a polygraph, a device which relies on an operator asking questions and measuring physiological responses like heart rate and perspiration, Converus’s technology measures “cognitive load” with an algorithm that processes eye movement.

Using cognitive load as a metric for lie detection relies on the assumption that it takes greater cognitive effort to invent a unique lie than to tell the truth. But the correlation between the cognitive effort recorded in involuntary eye movements and lying isn’t clear cut. Converus’s technology follows decades of research that has failed to identify a method that can reliably differentiate between stress and lying. As The Intercept reported in 2020, the shaky science underlying lie detection hasn’t stopped new technologies and methods from proliferating in law enforcement agencies across the country.

“If I asked you now, exactly one year ago today, what did you do? It’s probably easier to come up with a story than to really think about what I was doing,” Ewout Meijer, a professor of psychology at Maastricht University in the Netherlands, told The Intercept. Meijer, who studies lie detection technologies and the psychological theories underlying them, says that while some machines are accurate in reading physiological responses, those results don’t directly translate to accuracy at detecting lies. “There are boundary issues that mean this correlation between cognitive load and lying doesn’t always apply. But at least it is an alternative to the arousal-based approach,” he said.

Converus’s EyeDetect+, an upgraded version of EyeDetect, combines elements of the arousal-based assessment — which relies on traditional polygraph metrics like heart rate — with ocular measurement to increase accuracy, according to the company, which says EyeDetect+ can boast an accuracy rate of 90 percent. But Meijer says there are also problems with this approach. “The problem is not that we do not have the technology to measure emotions, we’ve had that for decades. The problem is in the underlying inference. If I measure emotion, what does that mean? It could mean that you’re lying. But it also could also mean that you are telling the truth, but you are really stressed and nervous about not being believed.”

“Comments by critics that the physiological changes recorded during polygraph and ocular-motor tests for deception do not directly correlate to truth-telling do not understand the meaning of ‘correlation,’” Warner told The Intercept. “If there were no correlation between certain physiological changes and a person’s deceptive status, it would not be possible to distinguish between truthful and deceptive people.”

In 2020, a member of Converus’s own advisory board expressed skepticism about the reliability of EyeDetect in an interview with MIT Technology Review. “I find the EyeDetect system to be really interesting, but on the other hand, I don’t use it,” he said. “I think the database is still relatively small, and it comes mostly from one laboratory. Until it’s expanded and other people have replicated it, I’d be reluctant to use it in the field.”

The scenarios for which Converus is marketing its lie detection tools are wide-ranging, according to the case studies presented as part of its sales pitch.

Life Renewal, a Christian counseling service in Texas, uses EyeDetect for marriage counseling. “The practice works with many sexual betrayal situations and — as is a common trend — lie detection is often used to help get to the truth so that healing can begin for the betrayed and the betrayer,” the Life Renewal case study reads. “With a lie detection solution, couples get to the truth more quickly.”

Another counseling center in Florida uses EyeDetect to address the challenges “with determining if unfaithful spouses or those in sex addiction therapy were truthful when disclosing the details of their sexual misconduct or sobriety in their therapeutic disclosure letter.”

Outside of therapy, the technology is being marketed for security applications including sex offender monitoring and truth-telling in border control scenarios. Converus also says its testing could be applied to intelligence scenarios related to espionage, terrorism, and trafficking. The company has presented at border security conferences and recently added Jayson P. Ahern, the former acting commissioner of U.S. Customs and Border Protection, to its board.

“We have not tested EyeDetect for every application in which it is used. However, in published research, we have established its generalizability across different ages, genders, educational level, languages, and cultures. Importantly, we compared ocular-motor changes in lab and field settings in a study with over 300 lab and field subjects and found no significant difference,” Warner said. “If results generalize across individual differences … and from the lab’s artificial and sterile conditions to the uncontrolled real-life conditions of the field, we are willing to conclude the technology is robust and can be applied effectively in a wide range of applications.”

Converus’s tests do not rely on a human operator but instead use an algorithm linked to monitoring components to assess truthfulness. Converus says that the lack of human operators reduces biases and creates a more accurate test. Leonard Saxe, a social psychologist at Brandeis University and a longtime skeptic of polygraph tests, says the operator is irrelevant.

“A machine that gives you a wrong reading, even if it’s a machine, is just as unreliable as a person that gives you a wrong reading,” Saxe said.

“It’s not a question of whether they’re telling the truth or not,” he added. “It’s a question for people at the border of whether they’re going to be admitted or imprisoned, and of whether they are scared or cognitively burdened with trying to give the right answer. For almost anybody being subjected to these tests, it’s an anxiety-provoking experience.”

Citing Saxe’s work, the American Psychology Association published a review largely discrediting polygraph testing in the early 2000s. “Most psychologists and other scientists agree that there is little basis for the validity of polygraph tests,” the review found. “Courts, including the United States Supreme Court … have repeatedly rejected the use of polygraph evidence because of its inherent unreliability.”

When asked why polygraph tests are still used in federal clearance screenings and hiring assessments, both Meijer and Saxe told The Intercept that the mere threat of polygraph testing can elicit pretest confessions and steer untruthful applicants away, regardless of the test’s accuracy. Whether the government should be relying on lie detection technologies as tools for truthfulness is another issue.

“When we think about ideal evidence-based medicine, we have a pyramid of evidence with case reports at the bottom and then meta-analysis of preferably randomized, double blind clinical trials,” Meijer said. “Based on that evidence, we decide what should be funded from public money and what shouldn’t. If you look at security, anybody who claims that they have a method that works gets government contracts, without this rigorous analysis. I find that weird, because this is public money, and also this is public money funding technology used in life-or-death situations.”