Rhetorical question? It's polygraph, drug sniffing dogs and breathalyzer. There was a court case a while back about drug sniffing dogs that presented the empirically demonstrated fact that the dogs present false positives all the time; but the judges concluded that it resulted in so many convictions that it would be a disaster to prevent their being evidence in court (analogous to witch dunkings, I suppose.) And breathalyzers are electronic devices whose source code is not auditable because the it is a "trade secret." Convicted by breathalyzer result? Can't challenge the black box's findings.
Also, how could sniffer dogs provide evidence for court? Surely they couldn't convict without actually finding drugs?
Probably DNA. If you suspect someone for various reasons, and then get a DNA match, it's almost certain that you have the right person. If, on the other hand, you have a DNA profile, and you search the 10,000,000 entry FBI CODIS database, there's somewhere between a 1 in 10k to 1 in 3m chance that you'll get a hit on someone other than the person who left the sample.
This seems to be rather understudied. Prosecutors like to talk about the probability that two people chosen at random would match. But that's not the correct value when you use DNA to locate the suspect.
One thing that would be useful would be for the FBI to report on collisions and near-collisions between known profiles added to the database.
I wonder what other type of crap is putting innocent people in jail? Maybe more interesting question: how can a jury of average citizens be expected to weigh this evidence when even FBI task forces can't completely figure it out?