The latest lie-detector trial suggests the best way to detect liars may be to lie.
Whether it be changes in body language, sweaty palms or fluctuations in brain activity, no surrogate marker of deception has yet proved reliable enough to be admitted as evidence in a court of law. Yet as governments and other large "shellers-out of cash know only too well, lies can also be expensive, which is why they continue to experiment with technologies that try to catch people out.
The latest is voice risk analysis (VRA), which, among others, Harrow Council in London is testing as a way of identifying fraudulent claims for social-security benefits. This week the council said that in the ten months from May 2007, when the system was installed, it had saved £420,000 ($828,000). VRA is not new. The version used in Harrow was developed by an Israeli company, Nemesysco, for use by the country's intelligence services to identify potential terrorists.
The system is said to operate by detecting changes in the sonic frequencies of people's voices that are caused by stress, and comparing these with a baseline created by their responses to unemotive questions, such as requests for their personal details. It places the changes on a spectrum of risk and, according to Capita, a firm which provided Harrow with the technology, operators then "form their own judgement of risk based on both the VRA technology and behavioural-analysis skills in which they are also trained.".
Mitchell Sommers, a psychologist at Washington University in St Louis, Missouri, with an interest in speech perception, says the findings confirm what other studies have shown about VRA: that it is not particularly good at detecting liars, but that it does act as an excellent deterrent. "These things do not work any better than chance," he says. Yet in America, insurance companies regularly claim that VRA helps them to reduce fraud by around 90%.
Harrow is happy that VRA acts as a deterrent, but they do not say that is all it does. Amir Liberman, the chief executive of Nemesysco, denies that it works by chance and says his company's own research shows it is highly predictive of lying even when people are not informed their voice is being analysed. The system does not measure lying itself, he adds, but the emotion associated with lying.