In response to the press attention on the use of voice risk analysis software (VRA) to prevent benefit fraud this factsheet examines how VRA is currently being used by local authorities as well as the Department for Work and Pensions' (DWP) plans for its wider implementation.
Whenever a customer provides information over the telephone VRA technology can detect any change in that caller’s voice whilst he or she is being questioned. The technology (software) tells a member of staff whether these changes indicate that the caller's responses are high risk or low risk. High risk is considered to indicate a possible fraudulent claim, which merits further investigation.
VRA technology looks for changes in voice patterns such as any hesitation or unwillingness to answer a question bur before it can do this the caller/staff member needs to first 'calibrate' the software at the beginning of an interview.
He or she does this by asking basic questions confirming the interviewee's personal details, such as his or her name, address and date of birth. This establishes the 'normal' pitch and tone of the interviewee's voice so that any change can be detected when the more probing questions begin.
When more detailed questioning begins if the claimant’s responses are hesitant or repetitive HR (high risk) flashes on the questioner’s screen. If not LR (low risk appears).
The ‘success’ of the system depends on three things:
The questioner has the power to override HR messages to prevent false readings - for example if someone is hard of hearing and hesitates or asks for questions to be repeated.
VRA is not new. It is used in the insurance industry as part of what is known as Advanced Validation Solutions (AVS), developed by a company called Digilog. AVS uses VRA technology.
The first use of VRA for benefit claimants was by Harrow Council, in May 2007, for housing benefit and council tax benefit (HB/CTB) claims. Using funding from the Department for Work and Pensions (DWP), Harrow Council began a year-long pilot.
It used Digilog’s Advanced Validation Solutions (AVS) in its call center. Digilog are working in partnership with Capita, who handle a number of government contracts as well as the congestion charging scheme in London.
At Harrow, claimants are sent an initial letter, warning them to expect a phone call in relation to their claim for benefit. This call could be to verify an existing claim or to make a new claim.
Customers can opt out of the call if they wish to do so. During the initial pilot over 5% (53 people) have done this in Harrow.
If the customer is low risk he or she is fast tracked through any claims process. If the call is to verify a claim no further action is taken.
Those whose responses are termed high risk are usually given a home visit. In some cases this may also later involve a fraud investigation.
Harrow Council claims that VRA has saved taxpayers £50,000 per month (or £300 per working hour). It also claims that out of nearly one thousand people to have been assessed using this new technology, 43 (4.3%) were labelled as ‘high-risk’ and their benefit claims were later found to be false and potentially fraudulent.
The council believes there is ‘strong anecdotal evidence’ that the lie detector test is having a significant deterrent effect on ‘would-be fraudsters’ as claimants were now volunteering information to say that their personal circumstances had changed and they no longer needed benefits.
Harrow also claim that, in addition to combating fraud, the aspiration of the original Harrow pilot was to replace their 30-page housing benefit review form and speed up their three to six week processing time with a 15-20 minute telephone call whilst also reducing the verification process.
Capita have also emphasised that VRA produces speedy claim processing and improved customer service (see the press release Voice risk analysis test to be piloted in more regions published on 5 August 2008 on the Capita website).
Following Harrow's pilot, six other councils began to pilot Digilog’s AVS solution and VRA technology as well as Harrow, with the first beginning in September 2007. These were:
In April 2008, the then, fraud minister James Plaskitt announced funding for a larger trial of VRA. These pilots, like the others, will be partnered by Capita. The following councils were successful in their bid to run a pilot:
The first of these new pilots began on 18 August 2008. All of the pilots were up and running by the end of November 2008. The pilots will run for 12 months, after which there will be an evaluation of the effectiveness of VRA technology in reducing fraud and error in HB. The original seven pilots will also continue.
In its 2005 strategy document Reducing fraud in the benefit system: achievements and ambitions, the Government made a public commitment to test Voice Risk Analysis technology, as one method of preventing fraudulent claims to Social Security benefits over the telephone.
The DWP is in the early stages of evaluation VRA. Eventually it is hoped that VRA will be introduced into local jobcentres. The DWP have identified three criteria crucial to the success of VRA:
Representatives from Jobcentre Plus have said that initial tests had concluded that the system could work with current Jobcentre Plus IT systems, that the operators could successfully use it, and that the tests did not adversely affect customer service standards.
The system is now being evaluated to establish how accurate the software is at identifying high and low risk calls/claims based on trials by local authorities and Jobcentre Plus.
So far DWP has looked at the data from six of the original seven local authority pilots (Warwick returned too few calls for analysis) and one pilot by Jobcentre Plus.
In each pilot a selection of customers were given a face to face review that aimed to corroborate the operator's judgment when using VRA. The results published in Hansard 11 Mar 2009 : Column 19WS do not appear conclusive.
It is hoped that the results of second phase of local authority pilots that are now underway will provide a broader evidence base and a clearer indication of VRA capacity to distinguish reliably between high and low risk cases. Results will be available in spring 2010.
Disability Alliance, together with other organisations, has already expressed some misgivings about the use of VRA for people with certain disabilities.
The DWP has indicated its willingness to engage in discussion about these issues and concerns. Presumably this will be some time after the initial evaluation of VRA. As and when this happens Disability Alliance will be involved. As other news breaks on this subject it will be posted on our website and this factsheet will be amended.
You can find out more about voice risk analysis and the Advanced Validation Solutions system on the Digilog at www.digiloguk.com. The website also includes an archive of press reports on VRA. There is also some information on VRA on the Capita website at www.capita.co.uk.
You can view information about the VRA local authority pilots in Housing Benefit Direct 83, available on the DWP website at www.dwp.gov.uk/housingbenefit/user-communications/.
The TUC have produced a report on VRA entitled lies, damned lies and lie detectors: introducing lie detector tests available from their website at www.tuc.org.uk.
There is also a useful article on VRA entitled Listening for Lies by Ryan Joe which was posted on www.speechtechmag.com on 1 Nov, 2007
You can obtain copies of this and all our factsheets by contacting Disability Alliance on 020 7247 8776 (voice and minicom) or by fax on 020 7247 8765.
www.disabilityalliance.org - 12 March 2009
