Sunday, 2 September 2007

Stressed benefits claimants must jump another hurdle

An article in the Observer today reports that Harrow Council has piloted the use of Voice Risk Analysis software in its benefit offices and found that, after its introduction "126 benefit cheats [have been caught] in just three months, saving [the] local authority £110,000".

The VRA software "enables trained operators to identify suspect cases at the start of a claim, helping to keep fraud out of the benefit system". It apparently does this by means of "thousands of mathematical calculations, resulting in the identification of different categories of emotional content."

A device that makes so many mathematical calculations naturally arouses our respect, and
you are probably now thinking "Wow, let's roll out this technology across the country." Before we do, it's worth asking the question "What evidence is there that this technology can detect people in the act of lying?"

The answer so far seems to be "Not a lot".

The use of electronic devices to analyse the human voice has a venerable history dating back at least thirty years. A range of industries, from insurance and banking to the legal profession claim to use some sort of voice analysis to carry out their work. There is an obvious law enforcement value in using such a device to detect when people are lying, and this is the reason so many police departments, particularly in the US have got in on the act. Despite this widespread adoption, a 2002 report "Voice Stress Devices and the Detection of Lies", published by the International Association of Chiefs of Police (IACP) National Law Enforcement Policy Center concluded that the accuracy of devices then on the market was "modest to poor".

Another test of two devices: the Vericator and the Diogenes Lantern concluded that
"After reviewing the three technical tests performed, it could be stated that these two VSA units do recognize stress. Although these systems state they detect deception, this was not proven."

To be fair to companies licensing voice analysis software, the IACP report rightly said that the dismal showing they'd seen so far did not preclude the invention of an effective "lie-detector" in the future. So, which system is Harrow Council using to catch its benefits fraudsters?

Well, the actual software Harrow uses has not been disclosed, but a little tracing reveals that Harrow Council uses Capita as its IT partner. A quick search of Capita's website reveals a press release from 2004 announcing the acquisition of Brownsword, a claims investigation service that came with a ten year license of "Advanced Validation Solutions".

This is a package that includes "Voice Risk Analysis" software developed by a firm called Digilog, who in turn license the core technology from Nemesysco: an Israeli company who claim that their algorithms underlay products like the Vericator and The Truster, the first of which fared so badly in the tests described above.

Interestingly, Nemesysco also market a "Love Detector" for both Pocket PC and a version that works with Skype.



I would therefore make an educated guess that Harrow is using software that comes originally from Nemesysco. Nemesysco say that its pre-2002 technology (discredited for the purposes of detecting falsehood in speech) is obsolete, and that its latest offerings are much more the ticket. The trouble is, I can find nothing in the scientific literature that attests to its fitness for purpose. This means that Harrow Council, and soon the Department of Work and Pensions, could well be detecting and prosecuting fraudsters using software whose only claim to be efficacious is that the company that makes it says it is.

2 comments:

Concerned Citizen said...

Sign a petition against it here
http://petitions.pm.gov.uk/ofbenefit/

remember to click the link in the confirmation email otherwise your name will not be registered.

Thanks

Sceptiphreniac said...

Thanks for taking the time to read my article. I agree with your aims, but value my privacy. For this reason, I will not be signing the petition.