Taxman rejects 'lie detector' tech
'Nonsense' ahead of Clegg speech
The government has denied claims it will extend use of telephone "lie detector" tests to the tax system.
Proponents of such software - known as voice risk analysis (VRA) - say it is able to calculate the probability someone is lying over the phone by measuring variations in their voice. Scientists charge it is no better at identifying fraud than tossing a coin.
The Daily Mail and the Telegraph both carried claims this morning that VRA would be used to identify tax cheats during telephone interviews with HMRC investigators. The reports came ahead of Nick Clegg's speech to the Liberal Democrat party conference, which will call for a crackdown on tax evasion.
However, an HMRC spokesman insisted the reports were "nonsense" and that there are no plans to use VRA in the tax system.
"HMRC has a responsibility to ensure that everyone pays there fair share of tax, in full and on time," he said.
"However, there are currently no plans for the department to use lie detector or voice detection software."
The Labour government began trials of a Capita VRA system in the benefits system in 2005, with "varied" results. In Wealden in Sussex, for example, claimants classified as "high risk" by the VRA system were only two per cent more likely to have their benefits changed after further investigation than those classified as "low risk".
Nevertheless, earlier this year the Department for Work and Pensions announced the trials would be extended to cover 24 local authorities.
Such enthusiasm for the technology also ignored a 2007 study in the International Journal of Speech, Language and the Law, which said that VRA vendors' claims amounted to "charlatanry".
"Our review of scientific studies will show that these machines perform at chance level," the authors, two language science professors, wrote.
"No qualified speech scientist believes in this nonsense," they concluded.
Nemesysco, the Israeli software company behind Capita's product, threatened to sue the journal for libel, so it withdrew the paper. It is still available online. ®
Sponsored: Hyper-scale data management