Speaking: the truth

I have a certain fondness for people who make remarkable technological claims. They appeal to my sense of the dramatic, my addiction to proper science fiction, and my journalistic laziness. If Intel says its latest chip is seven percent faster on interrupt handling, it's a lot of hard work to find out whether it's true and, at the end of it all, you don't really get a feeling that the world has made another leap towards joining the Galactic Federation. (Apologies to the engineers for whom reducing stack latency does feel like inventing warp drive.)

But if someone claims to have rewritten the laws of physics – Steorn, xG – then it's dead easy to check whether it's true or not. You just have to see the thing working on a desk in front of you. It's not my fault as a journalist if such companies go to great lengths to avoid such simple tests, and not your fault as a reader if you draw your own conclusions.

Which is why I am delighted to hear about yet another set of truly amazing claims, this time from a company called Nemesysco. This outfit makes voice analysers that are claimed to work so: “By utilizing a wide range spectrum analysis to detect minute involuntary changes in the speech waveform itself, LVA can detect anomalies in brain activity and classify them in terms of stress, excitement, deception, and varying emotional states,”

Mind reading through spectral analysis! W00t! Sign me up!

There is a big difference between Nemesysco and the other companies I mentioned: Nemesysco is happy to sell you the kit. Indeed, it has done so to our very own Department of Work and Pensions, which is keen to use the technology on benefit claimants. At a reported $25,000 a unit, I'd be keen to flog it to large organisations too, especially when it looks like a $2000 laptop and some software.

The trouble starts when one looks a bit more closely at the technology. Here, I can do no better than refer you to the Ministry of Truth, a fine blog which has done most of the spadework on this story.

They report, as far as I can tell entirely accurately, that a pair of Swedish researchers expert in the field of voice analysis – one the Professor of Phonetics at Stockholm University -- had a good look at Nemesysco and published a paper in a technical journal. They made some interesting claims. Things like the whole scheme relying on roughly 800 lines of 'ad hoc' Visual Basic, which you can inspect as they're part of the patents behind it They also don't pull any punches about the science behind the device - not, they said, that they could find much.

You can't read the paper online, unfortunately. The man behind Nemesysco promptly threatened the journal with libel, and the journal pulled the paper. Unfortunately, we have yet to read any refutation of the facts reported in that paper as none has been forthcoming. Doubtless soon, eh?

Meanwhile, taxpayers' money is being spent on this stuff and, worse, it's being built into the way the state assesses people in need. The Ministry of Truth is very keen to find out from the officials responsible exactly what's going on here, and I'm very keen for them to do so.

There is a much wider debate to be had about the utility of such devices and, indeed, all 'lie detector' technologies. As far as I can tell, none of them work reliably – although they have some theatrical value with the credulous. Here, however, the claims of the Swedish researchers are much simpler: the device simply cannot work as advertised, not because the science is controversial but because there is no coherent science at all.

To respond with a gagging threat rather than proof is in itself terribly interesting. Could it be that after all, like Steorn and xG, Nemesysco is also very unhappy with the idea of formal scientific analysis? I look forward to hearing more.