News for hackers. Stuff that matters.
Re:America's 250th DNA exoneration: How often is wrong perso... by ellyagg on Tuesday February 09, 2010 @12:58PM (#1112749)
> The certainty of DNA testing means we can be positive the 250 cases listed in the Innocence Project report didn't commit the crimes for which they were convicted

This is simply not true. Matching DNA against the defendant is basically 100%, but that doesn't take into account the circumstances of the case and where that DNA fits into the evidentiary chain. All the DNA has to do is provide reasonable doubt. It does not prove innocence. Indeed, contrary to the claim, it is virtually certain that some of the convicted in the overturned cases did commit their crimes.

Because of the political will behind resurrected DNA cases, the circumstantial link doesn't need to be very strong to have a case overturned via DNA.

Re:America's 250th DNA exoneration: How often is wrong perso... by barrkel on Tuesday February 09, 2010 @01:03PM (#1112761)
And of course, there is always Bayes to take into account. If a jury hears that there's only a "one in a million" chance that a DNA match would happen randomly, they may overweigh this fact, when put up against e.g. a city of 3 million people - which, assuming no other evidence and that the criminal is known to be from the city, would make the chance of this guy being the guilty party only 1 in 3.
Re:America's 250th DNA exoneration: How often is wrong perso... by btipling on Tuesday February 09, 2010 @01:56PM (#1112902)
Isn't it true that using Bayesian inference one should conclude that given a positive DNA match that it is more likely you have a false positive than an actual match? That isn't there are greater likelihood that there was a mistake in the DNA testing result than you actually find a match? :/
Re:America's 250th DNA exoneration: How often is wrong perso... by gjm11 on Tuesday February 09, 2010 @07:10PM (#1113787)
If the person who was tested was chosen at random from the population, then yes. But -- and this is an important point that somehow almost gets omitted when the usual examples of DNA evidence, medical tests, etc., are brought up -- when you're testing someone's DNA against samples from a crime scene, or testing whether a person has a rare disease, it's almost always because you already have evidence that brings Pr(they did it) or Pr(they have the disease) way above the baseline level. And in that case Pr(they did it | DNA evidence), or Pr(they have the disease | test result), may well be quite large.

(Of course that's not true if the person with matching DNA was found by some kind of search of a Vast DNA Database, or if the positive result on the medical test was the result of a mass screening exercise that tests everybody.)

Re:America's 250th DNA exoneration: How often is wrong perso... by Estragon on Tuesday February 09, 2010 @01:15PM (#1112796)
That reasoning would only hold if you surveyed the DNA of everyone in that city, which is never the case.
Re:America's 250th DNA exoneration: How often is wrong perso... by barrkel on Tuesday February 09, 2010 @01:39PM (#1112852)
No, you don't need to survey everyone in the city. Assume you instituted a DNA search, whereby you went from district to district and tested DNA until you got a match, the chance would still be only 1 in three, under the same assumptions.

And government policy here in the UK seems to be to retain the DNA of everyone that comes into contact with the state, so I expect it's only a matter of time until there will be chance collisions resulting in incorrect convictions.

Re:America's 250th DNA exoneration: How often is wrong perso... by Estragon on Tuesday February 09, 2010 @11:45PM (#1114236)
OK, it only applies if you survey everyone in the city until you get a match. Same difference. Yes, things may be heading in that direction. That was not the original argument, though.
Re:America's 250th DNA exoneration: How often is wrong perso... by eru on Tuesday February 09, 2010 @01:42PM (#1112865)
And if the DNA was every hint you had.