Monday, 23 February 2009

How not to do science

From a scientific point of view, this is about as wrong as you can be.

The journal Science last week had a story that on the surface looks like a minor spat between scientists but in reality has several unfortunate consequences. According to Science, an Israeli company (Nemesysco) is selling a device using Layered Voice Analysis (LVA), which they say can analyse voices to help determine whether the speaker is possibly being deceitful. They have a patent on the technique and have published some papers on the method of analysis. But they also claim that their device implements more than they have published because they think they have been ripped off by publishing too much in the past.

Enter Anders Eriksson and Francisco Lacerda, two researchers in phonetics in Sweden, who have studied this device and came to the conclusion that it was not all it was cracked up to be, reporting their results in December 2007 in the peer reviewed publication The International Journal of Speech, Language and the Law under the inflammatory title of Charlatanry in forensic speech science: A problem to be taken seriously. A founder of Nemesysco, Amir Liberman, took exception to what he described as a personal attack and asked his lawyers to contact the journal, who, being a small bi-annual publication could not afford any legal costs and pulled the article from their website. This led to the Science piece because Lacerda counter claimed that Liberman is attempting to stifle scientific development.

The LVA devices have been bought by 25 local administrations in the UK to assess callers with a view to further investigation for trying to defraud the benefits agencies. A spokesman for the Department of Work and Pensions said that a decision will be made in 2010 as to whether to roll it out nationwide. The figures quoted from the London Borough of Harrow are insufficient to judge whether they could have saved just as much if they took a random selection of applicants rather than the ones LVA picked out. There is no comment on whether they considered the effect of the tool on matters of privacy, liberty or legality. Thoughts of the polygraph come to mind.

So, what do have here?

First, a company that wants to keep information to itself so it can make money from it. This in itself is a warning to the sceptical that there may be more to this than meets the eye.

Second, researchers who have tried to assess the device, and written a tactless article. Even if their blunt criticisms are valid, this is not the way to go about influencing people.

Third, a journal that apparently took a legal risk which its editors could have avoided by insisting on a rewording of the disputed article.

Fourth, customers, most notably people in a position of power, who have taken something on the say-so of a sales force without having had the efficacy independently assessed, and apparently with no plan to do an adequate assessment themselves.

My point in writing this piece is not to take any sides in this particular matter since I do not have the full facts, but to ask what they all thought they were doing. Science can provide data, but only when access to it is unconstrained and open to sceptical checking. The reporters of experimental results have a duty to do so in an unemotional and straightforward manner. One part of the task of journal editors is to ensure papers are civilised and clear. And customers who have an opportunity to acquire good data to make a decision that can be supported in public, but do not seem to be doing so, are misusing the data they are collecting. The only gainers in this sorry state are the lawyers.

But everyone could have been better off if they had all been more conscientious. Customers could get better data for a more compelling decision; editors could avoid legal disputes; authors could provide uncontroversial data to make a point; and companies could gain through better co-operation with academia and their users.

I might have shrugged this off as a case of Homo sapiens behaving typically apart from the understanding that there is a chance that every applicant for support from the state may eventually be subject to a dubious attack on their good name by the use of this device.

If we want to attract more to the sciences, then we need to be rigorous throughout. This story is an example of how to get it wrong all the way through, from idea to use.

No comments: