Government abandons benefit-cheat lie detectors The government has abandoned plans to use a telephone-based lie detector to catch benefits fiddlers, following criticism by scientists that the technology is no more reliable than tossing a coin. Analysis of "Voice Risk Analysis" (VRA) software did not conclude it is effective, the Department of Work and Pensions said. "The …
So that's £2.16m wasted that could have been saved with 15 minutes web search and one phone call to one of the universities that the government is already funding?
Oh, and if they really had to waste millions, why couldn't they just burn the cash? I'm sure most taxpayers would prefer that to seeing Capita get it!
That depends which bluff you're trying to pull off:
a) Pretending that the software works in the hope some people may be more honest
b) Using having the software as a cover for basing investigations on tip-offs (partly shielding those giving the tip-offs)
c) Using having the software as a cover for basing investigations on profiling (shielding the council from accusations of discrimination/racism/whatever)
In case c), it's probably worth actually having the software for arse-covering in case anyone complains. In cases a) or b), turning out not to have it *could* be explained as a cunning bluff.
Though, of course, if you're not relying on it actually working, it'd be possible to buy (or pretend to buy) cheap software rather than expensive software.
"The best plan would be to spend exactly £0.00 on voice scanning technology, but then tell the public that you've implemented it anyway."
I'll bet that would work as well as dummy CCTV cameras, ie not at all.
This was clearly a dumb idea rooted not in the desire to catch more fraudulent claimants, but mainly in saving money. The fact that is was supposed to indentify high risk claimants demonstrates that the plan was to cut down on the number of claims which were investigated.
Those seem to work just as well as "working" CCTVs, only without the dubious benefit of function creep into fining littering and missing dog leashes, oh and the "occasional" govermental perving, of course. They do require a lot less gratuitous data retention though. Should be a major cost savings in backups alone. Where's the budget cutter man when you need him?
"At no stage did the evaluation carried out by the Department explicitly consider the effectiveness of the technological aspects of VRA."
Oh well that's OK then. Just keep commissioning systems that clearly won't work and we'll all just pay for it. Thinking ahead we could do with the following:
Chocolate teapots;
Bags of sparks for the grinder;
Virtual RAM for the server;
Everything else on this list http://www.messybeast.com/dragonqueen/fools-errands.htm
That really is a painful amount to spend on trials of snakeoil.
How much do we lose to benefit thieves a year anyway? I want to know how long this scheme would have taken to pay for itself, taking into account the cost of the trials and projected cost of a national roll-out, taking into account how effective it was expected to be at categorising people as thief/not-thief and how many claimaints it could be administered to.
Why do I also suspect that these were not so much trials in the sense that a randomised controlled trial is a trial (i.e. able to provide actual answers on whether it worked or not) but a trial in the sense that LET'S DO A THING AND SEE SOME OTHER THING! is a trial?
2.16 million...that's about 1.5 absolutely world-class clinical trials they spent on this.
There are simply too many signal degradations over any telephone carrier system to make these detectors accurate.
Whether 'landline' or cell all systems introduce phase distortion, flutter, audio frequency changes, etc. which are readily recognisable by telephone users.
Many countries have tossed out lie detector equipment, except for the U.S.A., as most courts recognise they are not accurate and their output requires 'interpretation' which can introduce bias.
These voice stress systems use similar principles. A pathological liar will always show 'true' readings as they believe their lies.
Lie detector technology does not produce any evidence which is admissible in court in the US either. The only way it is used anywhere with any legitimacy is essentially as intimidation and/or to produce possible leads. A failed lie detector test isn't even enough (by itself) to get a warrant to search as house.
Which is nice, since a lot of insurance companies at least claim they use this technology now. They certainly were using it on a tv show which concentrated on the 'fake accident' makers. Even showed us how the voice changed when lying...
Heaven forbid it was just another way for insurers to refuse a claim...
So why is your local authority using technology that doesn't work ? If you want to know whether to investigate someone (or not) tossing a coin would seem to be as effective - and probably fairer.
How many stressed out people have had their rightful benefits delayed because of a false-positive from your system ? How many cheats have got away with it because "the computer says ok" ?
*face*palm*
Install the lie detector in every microphone in both houses and especially the party press rooms. Hand the system out to news crews.
Make sure everyone watching or listening can see the result - like a really big red light and loud buzzer on the mic.
Now that I come to think of it - implant it in all top-level politicians. Connected to a contact taser, anally installed. They may switch it of in private but any official word they say will be tested.
After all, the system works and politicians would surely never lie to the public, wouldn't they? So what do they have to fear? Pants on fire?
It's not so much signal degradation on the line that's the issue, it's the plain simple fact that telephones are bandwidth limited to approximiately 4KHz, reduced from around 20KHz for normal human hearing.
There's so much information lost , but it's not a degredation, that loss is designed in from the very start.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
