Is voice risk analysis a sound investment?

The Department for Work and Pensions (DWP) is spending 1.5 million on voice risk analysis (VRA) software, despite an academic paper describing "the use of these machines as charlatanry" and early piloting results being too varied to draw conclusions from, according to a report in the Guardian.

The government is currently testing VRA - software used to detect if somebody is lying over the phone - as one method of preventing fraudulent claims to social security benefits. It is currently used in the private sector by car insurance firms stating it can detect false claims.

Last week, Minister for Employment and Welfare Reform Tony McNulty reported the results of such trials to the House of Commons as being "varied from pilot to pilot." A second phase of piloting, with a further 24 local authorities getting involved at the cost of 1.5 million, is now being undertaken and the results will be available in spring 2010.

A spokesperson from the DWP said: "The early findings from the VRA pilots confirm the early positive results reported in some local authorities, but reinforces the need to pilot further for more conclusive evidence."

One of the initial councils to pilot the scheme was Harrow, given just under 50,000 from the DWP to fund the project.

Paul Osborn, portfolio holder for performance, communication and corporate services at the council, said: "The Voice Risk Analysis technology is highly effective in identifying fraudulent claims over the telephone. Harrow Council has been using it to catch out benefit fraudsters who otherwise might still be at large."

He added: "Our success in using VRA is measurable in both identifying such people and saving money in the process."

Even with good recommendations, the government findings so far can neither confirm nor deny the scheme's validity and an academic paper has denounced the software, claiming it was "at the astrology end of the validity spectrum".

Francisco Lacerda, a professor of linguistics at Stockholm University, and Anders Eriksson, professor of phonetics at Gothenburg University, released their paper in Journal of Speech, Article Language and the Law called Charlatanry in forensic speech science: A problem to be taken seriously'. The paper questions why millions of pounds is being spent on a system that has not been scientifically proven as authentic.

Lacerda told the Guardian: "At best, this thing is giving you an indication of how [voice] pitch is changing. But there's so much contamination by other [noise] factors that it's a rather crude measure."

VRA software is designed by Amir Liberman from Nemesysco, who is based in Israel. Liberman claimed that the two professors never tested the software or ask any questions about it before publishing the paper.

He told IT PRO: "How can you testify the taste of an apple if you have never tasted it? I have spent 12 sleepless years researching and doing what I do. To call me a charlatan is like nothing I have ever experienced."

"We are a research company. What we do is research. We test over and over and over again. My reputation is more important to me than any profit I could make from this software."

In the UK, we will have to wait for the results of the further piloting to see the software's effectiveness but it seems the worldwide debate shall continue for some time.

Jennifer Scott

Jennifer Scott is a former freelance journalist and currently political reporter for Sky News. She has a varied writing history, having started her career at Dennis Publishing, working in various roles across its business technology titles, including ITPro. Jennifer has specialised in a number of areas over the years and has produced a wealth of content for ITPro, focusing largely on data storage, networking, cloud computing, and telecommunications.

Most recently Jennifer has turned her skills to the political sphere and broadcast journalism, where she has worked for the BBC as a political reporter, before moving to Sky News.