How Accurate is DNA - not as much as you think
I've written before about how DNA has been held up as the "gold standard" - the standard against which all other forensic evidence should be measured. As many expected, that declaration was a tad premature.
The problem is not with the actual DNA results. If done properly, it is what it is. Anyone who has seen DNA results knows there is a sequence of numbers, and graphs from the actual tests. The problem comes when a human being looks at that data, and interprets it. At that point problems can arise - including subjectivity.
Itiel Dror and Greg Hampikian recently published a paper titled Subjectivity and Bias in forensic DNA mixture interpretation. They wanted to address the potential problems in interpreting evidence when there may be multiple suspects. The situation most commonly occurs in sexual assault cases where there are multiple perpetrators. There may not be enough to conclusively identify one person, but there may be enough to say an individual cannot be excluded - which in the minds of most jurors means you must have done it.
They obtained an actual case out of Georgia that involved a gang rape. One of the alleged suspects identified and testified against the others. The results of the DNA analysis - by examiners who knew who the suspects were - could not exclude the others, which corroborated the co-defendant testimony.
The actual data was submitted to 17 qualified analysts who routinely do forensic work. No other facts were sent, so the examiners did not know who the suspects were. The results obtained without that contextual information were startling. Only one of the 17 agreed with the original examiner. Even more startling is that 12 examiners would have EXCLUDED the suspect they looked at. The remaining 4 would have called the results inconclusive.
The implications are clearly serious - the corroborating evidence didn't corroborate the co-defendant, but actually contradicted him. Even more troubling though is the unmistakable conclusion that subjectivity does in fact play a role - even if not a conscious one. The probability that this was the result of a simple mistake is simply not believable. Perhaps if the consensus was the data was inconclusive you might chalk it up to a difference in opinion. But when so many examiners read the evidence the exact opposite something more is at play
How - or even if - the scientific community addresses this problem is an open question. The lesson for lawyers is clear. You cannot simply rely on the state's experts, who performed their analysis with knowledge of the facts and who the police were trying to build a case against. You must have someone else look at the evidence, and if their results are different challenge.
As for those already convicted, that will have to be a subject for another post.
