Slashdot Mirror
Ohio Court Admits Lie Detector Tests As Evidence
An anonymous reader writes "Last month, an Ohio court set a new precedent by allowing polygraph test results to be entered as evidence in a criminal trial. Do lie detectors really belong in the court room? AntiPolygraph.org critiques the polygraph evidence from the this precedential case (Ohio v. Sharma)."
Imho, polygraphs should not be used. The simple reason is that some of the more violent and unpleasant people, psychopaths, show blunted responses in psychophysiological tests compared to your 'regular' violent perpetrator. As psychopaths tend to be the ones we should really keep off the streets then a misinformed jury might take polygraph results as definitive evident the perpetrator (psychopath) had not committed the offense and judge accordingly. Also, with a bit of practice and insight, some people are able to control their responses or give misleading results. There's no definitive objective means determining whether someone is telling the truth or not... next to honest evidence the polygraph is pretty useless. It's a nice idea but anyone who has used these psychophysiological tests will know, for every half decent result you also get a fair bit of noise (excluding, of course, the people ho make and sell polygraph tests).
I remember seeing a clip of Groucho being subjected to a polygraph from the 50s, and while funny, it didn't inspire any sort of confidence in the technology.
I agree, that considering how many minor things are consider to taint the jury, a polygraph is probably just about the worst of them. The reliability just isn't there, and even when they are accurate, they don't really give any indication of what the lie actually is.
Worse, they tend to work worse when the subject is already under stress. Overall, the technology just isn't there, and won't ever get there. If anything is more reliable, it'll be of a different form, probably something that scans the brain directly. Even that though is going to be tough.
The most common figure for the accuracy of polygraph tests is 70%. Which sounds reasonable, until you realise that since the situation is binomal -- i.e. the only possible results are "truth" and "lie", so pure chance (e.g. flipping a coin) would give you 50% accuracy; at which point 70% starts to look considerably less impressive.
As I understand it, the most useful (from the police's point of view) way to use of lie detectors is psychological: pretend that they're 100% accurate, get the suspect to say "I didn't do it", bluff and claim that "The Machine Knows You're Lying", and get them to give a confession that way. Of course, such a strategy will fail if the polygraph becomes so widely used that everyone becomes familiar with its limitations.
What's purple and commutes? An Abelian grape.