Friday, August 31, 2007

telling the truth about lie detectors

ABC's Charles Gibson had a story, repeated here, about a new device that purports to look at the brain to tell when you're lying. On the web article, it comes complete with a nifty comparison of a lying brain and a truth-telling brain.

Nifty photos, yay. But boo to the complete and utter lack of research into this story.

Basically this article took the press releases from the company that calls itself "No Lie MRI" and reported it uncritically. Absolute bull with less actual reporting than you'd see in an average junior high school newspaper. You want a REAL article on this with actual original research, facts, interviews etc, read Margaret Talbot's recent article in the New Yorker. Talbot actually talks to other researchers. Talbot actually acknowledges the facts that the testing protocols for such lie detectors (brain-based like the No Lie MRI or others) are badly flawed and don't, for example, include testing on people who are drunk or on other drugs, who are psychotic or mentally unstable. Talbot points out that a claimed 90% success rate is (1) unverifiable; (2) exaggerated by the fact that many people will confess when confronted by a lie detector test - including occasionally people subsequently verified by other evidence or confessions to be INNOCENT; and (3) leaves a 10% failure rate that is far, far too high and can result in liars getting away with it and much worse, innocent people being tarred with some crime. She points out that all the stuff that a polygraph (beloved by so many, and so very unreliable) measures is just basically a measure of STRESS, which could just as easily be caused by being an honest person telling the truth in a hostile situation (ie, on a lie detector!) as it would be by lying. SHE points out that the so-called visual cues for lying are NOT in fact universal - some people don't like to make eye contact period, whether they are telling the truth or lying.

In other words, Talbot's article was a fine piece of reporting. ABC's segment is a sweet piece of PR for No Lie MRI. Unfortunately, the Charles Gibson "report" will be seen by probably 100 times as many people as will read the New Yorker article. And the lies about allegedly being able to mechanically detect lies will be perpetuated, to the benefit of those most interested in making a buck or closing a case.

This was a shoddy piece of journalism, even by mainstream media standards. No lie.