The Pentagon is ploughing $30.3 million into an AI-powered lie detector called Polygraph+, hoping to refine the notoriously unreliable polygraph test. The project, set to run for five years, will use new sensor technologies and machine learning to improve the accuracy of lie detection. Experts, however, remain sceptical, warning that even with AI, the science of deception detection remains unreliable and fraught with issues of bias and false positives.

The move comes amid heightened tensions within the Department of Defense, where polygraph tests are increasingly being used to detect leaks. In 2023, the Defense Innovation Unit (DIU) ran an open submission process to find companies that could develop deception detection technologies, selecting Presage Technologies and Altec Research as potential partners. Both companies are working on non-contact sensing technologies, including head movement and facial temperature tracking.

Despite centuries of effort, no reliable lie detector has been invented. The American Polygraph Association claims the polygraph is between 80 and 94% accurate, but research suggests humans can spot a lie only about half the time without any technical assistance. Even with AI, the polygraph's biggest impact is often deterrent, making people confess before the test even begins. The problem, as Sophie van der Zee, an associate professor who studies deception, points out, is that there is no single, reliable sign of lying that works for everyone.

While AI could potentially improve the polygraph by finding patterns in the data, the inherent unreliability of polygraphs means that false positives and negatives remain a significant risk. The DoD employs 2.8 million staff, and an imperfect system at that scale could lead to thousands of wrongful accusations. Critics argue that the focus should be on improving the overall security and vetting processes rather than relying on unreliable technology.

Ultimately, the Pentagon's aspirations for Polygraph+ may be seen as a misguided effort to modernize an already flawed system. As Kyri Kotsoglou from Northumbria Law School puts it, it's a 'misguided effort to reduce the complex to something tangible'. Whether or not Polygraph+ will succeed, the debate over the reliability of lie detection technology will continue.