This problem is bad, and it is, at its heart, a money issue.

Maybe the biggest problem with forensic evidence is that defense attorneys don’t challenge it in court, because they themselves tend to believe it’s real science (as perhaps they are taught in law school).

Having said that, I have to agree with Michael Chaney. The justice system and the public are always resistant to anything that would impede their ability to get convictions.

Until that is done, I will regard all ‘forensic science’ as little more than 21st century phrenology.

1. Money for forensics should be equally split between prosecution and defense.
2. Forensics lab should have absolutely no direct contact with lawyers, and ultimately the technician who performs the test should not know what crime it’s connected with, who the client is (prosecution or defense), or anything else beyond the absolute minimum requirements (e.g. Here’s a fingerprint lifted from the scene and here’s a suspect’s fingerprint, see if you think they match).
3. All tests should be scientifically conducted using double-blind studies where possible.
4. Any evidence not disclosed in advance to the defense should be disallowed in court.
5. All studies should be filmed with the film shown to the jury. Studies that can be conducted in the courtroom should be. Example: “this money should be seized – my dog hits on it. Fine. Bailiff collects and tags random money samples from around the court room, numbers each, dog and handler are then brought in to find which of the 10 piles of money are the one. Bailiff leaves after setting up the experiment but before dog and handler come in. After dog and handler are gone, bailiff returns and reveals which pile was the one. The jury can then see if the dog is reliable. Dog hits on more than one pile of cash means he’s not reliable and should be decertified.” Honest prosecutors would be happy with such an arrangement.

As Radley said, though, they’re not looking for legitimate science – they’re looking to pin the crime on somebody.

Of course this kind of thing is so far over the head of most lawyers it is rather pathetic. You’d think that people who have to weigh the merits of evidence would have at laest a rudimentary understanding of how to do that via probability theory, but math is hard for most people and when you have to protect that GPA to get into law school…..

Structural reforms to address such problems of rational bias include independence from law enforcement, blind proficiency testing and separation of test from interpretation.

Stating one’s prior probability that the defendant is guilty would also work like a freaking charm. It is explicit and obvious and no way to dodge it.

You can never remove the human component from scientifc inquiry; even if you use a computer to scan, say, DNA “matches,” and sort them into a “definite” pile and a “dubious” pile, some person or team has to write the programs and set up the instrument to sort the probable matches. However, Radley makes an extremely wise point about blind checking. It’s almost never the case in the scientific world that a person with a vested interest in the data saying X will argue vehemently that the data say Y, unless it’s absolutely unambiguous. On the other hand, scientists with no vested career interests in X are much more likely to say “Hey, this clearly shows Y; come on.” So while human biases can never be removed from the human endeavor of scientific modeling, they can be greatly mitigated by separation of the police hypothesis from the forensic determination. This is similar to the notion that line-ups should be done in a fashion where there is minimal opportunity for accidental or purposeful police interference.