24 Jan 2013

Voice risk analysis software attacked

The leader of the Conservative group at Cornwall Council has sensationally quit her role on the Cabinet in a row over the use of lie detector tests on people claiming single person's council tax discount.

Fiona Ferguson claimed in her resignation email – which was placed online by councillors – that leader of the council Jim Currie had threatened to sack her from her Cabinet post if she revealed the details of the lie detector test being used.

The Voice Risk Analysis (VRA) software – which is known as a lie detector test – is allegedly being used by company Capita which has been employed by the council to carry out a review of people claiming single person's council tax discount.

The relief allows anyone living on their own to claim 25% of their council tax bills. Letters were sent out to those claiming the benefit last year and these were being followed up with telephone interviews.

During these interviews, Mrs Ferguson claims, the company was going to use VRA.

On Capita's website it states that VRA is "capable of identifying stress and emotion in a caller's voice pattern".

In her resignation email to Mr Currie she states that VRA had previously been reviewed by the Department of Work and Pensions which found that it was not reliable.

She wrote:
I have discussed this matter with the monitoring officer. He has advised me that, as this is an operational matter in relation to a contract that the council has already entered into, he strongly advises me that I should not require that this software is not used. If, contrary to his advice, I maintain my stance that we must not use this software then officers will comply provided you also agree.

You have made it clear to me that you will not agree. Indeed, you have said that I will be 'sacked' if I inform members that this software will be used.

That will not be necessary. Please accept my resignation with immediate effect.

May I say that I have no reason whatsoever to believe that you were aware of this aspect of the contract before I drew it to your attention. I also appreciate that you are in a difficult position in view of the monitoring officer's advice.

But I do not believe that his advice is correct and I cannot accept it on ethical grounds. I also do not believe that it will help the council to pursue fraud (which we must surely do) if the public think we are using this software.

Finally, I fear that it will be extremely damaging to our reputation.

Therefore, I am launching a petition to require any use of this type of technology to be approved by full council.
In a statement Cornwall Council said that the interviews were being carried out to verify details provided by claimants.

It said: "These interviews will be carried out by trained assessors who will use specialist technology to assist this process. This technology has been successfully used by a number of local authorities carrying out similar reviews. All claimants will be advised that the calls will be recorded, monitored and used for fraud prevention purposes."
  • What is the problem here? It's not as if the software makes decisions. It just suggests which cases might be investigated further.

    Council tax single person discount fraud costs between £60m and £300m, depending on who you believe - the DWP figure is the lowest, of course.

    Why would VRA software be unacceptable in principle?


Anonymous said...

Only works if you have to have verbal contact with them. That's if it works at all

John Page said...

Well they're following up letters with phone calls.

Anonymous said...
This comment has been removed by a blog administrator.
Unity_MoT said...

Why would VRA software be unacceptable in principle?"

Well, for starters, because it doesn't work. The system itself is nothing more than fraudulent piece of pseudoscience with no basis in any kind of credible speech science.

I have all the data from the DWP trials (via FOIA) so let's lay it out for you.

The official estimate for the level of fraud in single person discounts is 4%,ie 4 fraudulent claims in every 100. If you were to assess 100 claimants with this system then, based on the trial results it would correctly pick up two false claims, miss two entirely and wrongly identify 22 claimants as potential fraudsters, even though they gave accurate information - and the system would perform go that level only 6% of the time.

It doesn't work and Capita are only still pushing it, and pretending that it does because they paid £8.5 million to buy the company the had exclusive rights to the system and are desperately trying to get their money back on a total lemon.

John Page said...

Use the blog comments to campaign against the blog's views by all means, but not anonymously.

The anonymous posting facility is for benefit system employees. Other posters are expected to say who they are.