4 ms·

> 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 convic

by ellyagg 17y ago

> 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.

barrkel 17y ago
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.
- Estragon 17y ago
  That reasoning would only hold if you surveyed the DNA of everyone in that city, which is never the case.
  - barrkel 17y ago
    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.
    - Estragon 17y ago
      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.
  - eru 17y ago
    And if the DNA was every hint you had.
- btipling 17y ago
  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? :/
  - gjm11 17y ago
    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.)

by ellyagg 17y ago

barrkel 17y ago

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.

Estragon 17y ago
That reasoning would only hold if you surveyed the DNA of everyone in that city, which is never the case.
- barrkel 17y ago
  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.
  - Estragon 17y ago
    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.
- eru 17y ago
  And if the DNA was every hint you had.
btipling 17y ago
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? :/
- gjm11 17y ago
  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.)