Slashdot Mirror
On the Dangers and Potential Abuses of DNA Familial Searching
Advocatus Diaboli sends a story of how a high tech forensic procedure almost led investigators to the wrong person. In 1996, a young woman named Angie Dodge was assaulted and murdered in Idaho Falls, Idaho. There was a conviction in the case, but later reports claimed the wrong man was in prison, and police thought there were more than one attacker anyway. This eventually led to the re-opening of the investigation. Using DNA evidence that had been preserved from the crime scene, police used a controversial technique called familial searching to try to find a lead. This method is used when there is no direct DNA match within the available databases. Instead, it tries to identify family members of the suspect. Police found a partial match, which eventually led them to Michael Usry, a New Orleans filmmaker. They convinced a judge to provide a search warrant to extract Usry's DNA and test it against the sample. It wasn't until a month after the extraction that they told him he'd been cleared.
No mistake was made. The police checked a potential suspect and cleared him.
Well, this should cause anyone using a DNA service or donating their DNA to science to think twice:
The elder Usry, who lives outside Jackson, Mississippi, said his DNA entered the equation through a project, sponsored years ago by the Mormon church, in which members gave DNA samples to the Sorenson Molecular Genealogy Foundation, a nonprofit whose forensic assets have been acquired by Ancestry.com, the world’s largest for-profit genealogy company.
What the actual fuck?
My understanding is that DNA matches are generally not accurate enough to be accepted by scientists as a conclusive match, though perhaps the techniques have improved since then. That's actually a big part of the problem with them - people believe them to be considerably more accurate than they actually are. Maybe they're even 99% accurate, but that would still mean that there's a 1 in 100 chance that you'll come up as a false positive: enough to worry about I'd say. And if the accuracy isn't even that good... well it makes DNA databases being used for identification a bit worrying. Not so much because they're not useful, but because the police, lawyers, judge, and jury all tend to believe that a match is far more conclusive than it really is. Especially the jury, whose familiarity with DNA evidence comes from daytime television, where it *always* fingers the guilty party.
Basically nobody involved is likely to have the grasp of statistics necessary to properly parse the results - even doctors, who you would expect to deal with false positives/negatives all the time often get it wrong. Test is 90% accurate, that means if you come up positive there's a 90% chance you've got the disease, right? Wrong. You also need to factor in the independent chance of actually having the disease. Say there's a 2% chance that a person chosen at random is infected: that means out of 100 people you will (on average) have 10 false positives and two actual positives: testing positive only means there's a 2/12 (17%) chance you're infected.
Now consider DNA evidence - let's be generous and say it's 99.999% accurate: there's 400 million people in the US, so you can expect there will be 400,000 false positives, only one of which actually be guilty. Me, I don't like those odds.
--- Most topics have many sides worth arguing, allow me to take one opposite you.