THE UK government has dropped plans to introduce lie detector tests to catch benefit fraudsters after trials found that the technology was not sufficiently reliable.
The department for work and pensions has given up on “voice risk analysis” (VRA) software after spending £2.16 million on trials to assess whether the technology can identify people who are trying to fiddle the system when it eavesdrops on their telephone calls to benefit offices.
Plans to install lie detectors were hailed as a vote-winning move to get tough on benefit cheats when they were unveiled by former prime minister Gordon Brown in December 2008.
VRA is meant to detect signs of stress in a caller’s voice by analysing short snippets of speech, but critics say the system is not powerful enough to distinguish cheats from honest callers.
In 23 pilot studies, local authorities used the lie detector system to analyse phone calls from people applying or updating existing claims for housing benefit (for rent), council tax (for local property tax), income support (for people on low incomes) and jobseeker’s allowance (for unemployed). The technology was judged a success in only five of the trials.
A spokeswoman for the department for work and pensions said: “We have got the analysis back and have been going through whether it works when applied to the benefits system. This is the first time it has been used in the benefits system and the decision is that it is not very good value for money,” she said.
The department organised two groups of trials. The first, in 2008, cost £460,000 and involved six local authorities and the department’s executive agency, Jobcentre Plus. The second-phase trial was expanded to 24 local authorities at a cost of £1.7 million. Information from 45,000 calls was included in the evaluation, the department said.
Nine local authorities trialled the lie detector on calls about new benefit claims. Of these, only three said it worked well enough to highlight risky callers without raising too many false alarms. Of 12 local authorities who used VRA to spot cheats during benefit reviews, only one judged the trial to have been a success.
Voice risk analysis has been mired in controversy since scientists raised doubts over the technology soon after it reached the market. In 2007, two Swedish researchers, Anders Eriksson and Francisco Lacerda, published their own analysis of VRA in the International Journal of Speech Language and Law.
They found no scientific evidence to support claims for the device made by Israeli manufacturer Nemesysco.
Mr Eriksson and Mr Lacerda went on to say the software was “at the astrology end of the validity spectrum”.
Following complaints from Nemesysco’s founder, the article was withdrawn from the website of the journal’s publisher.
Mr Lacerda, who is head of phonetics at Stockholm University, said: “My only surprise is that it didn’t come earlier. There is no basis for the device at all, so I would be surprised if they had reached another conclusion.” – (Guardian service)