The whole field of forensics is looking a lot like junk science. I remember when fingerprints were exposed as bogus. Now it seems every other type of forensics turns out to be faulty when it has the slightest scrutiny.
Even forensic DNA testing while potentially good science has ridiculous claims made about it - "one in a billion chance of being wrong". But it turns out they wrongly convict people using "unimpeachable" DNA evidence fairly often [1].
There's a general principal involved, that once you start throwing round numbers like 'one in a million' around, you're almost certainly overlooking larger sources of error. Eg, someone fucking up in the lab is much more likely than one in a million. There was a famous case where they were looking for (what they thought was) a serial killer who showed up in a bunch of DNA tests, and it turned out to be some factory worker who had contaminated the swabs.
I think of it as the "noise floor" of probability. It's great that, say, the odds of a DNA match is one in a trillion, but the odds of someone getting there wrong sample, misreading the results, deliberately forging the results, etc. is so much higher that the probability really only tops out (or bottoms out, depending on your perspective) at the noise floor.
Comes up a lot in our world because computers so easily generate those large probabilities by being the things that can run millions of "samples" of some distribution per second, but you always have to cap those analyses to the probability noise floor or you'll be very misled.
This is basically Bayes rule: the probability we have an accurate test given perfect conditions.
But what is the baseline rate of perfect conditions? If it's less than 1, then our "posterior" probability will not be one in a trillion anymore, but something far less.
They don't stand up to the claims that prosecutors make about them and get treated as far far more reliable evidence than they actually are.
There is some signal in there but analysis of only a few prints or partial prints should be treated as extremely suspect if that's all you got. The place where the limited data is at its strongest is when you know there were only say 3 people who could have made that print because you can reliably rule the others out if they physically couldn't have made the print you lifted.
But if you're going fishing and all you have is some prints matched against a database the idea of a "match" meaning anything other than "is part of a cohort of unknown size that could have made that print" should be thrown out.
Quoting from [0], but there are many other sources with similar stats:
> One study of FBI fingerprint analysis found that false-positive errors in which a fingerprint was incorrectly matched with a suspect occurred in one out of every 306 cases. A study of a Florida police crime lab was even more troubling, finding that false-positive errors occurred in one out of 18 cases.
That is certainly very far from bogus, unless you use bogus language like "it's a match". If instead you reasonably estimate the probability of error, updating your prior by a test with 1/18 error rate is very valuable.
and that would be the point. Any single piece of evidence, particularly something with a noticable error rate, should not be enough to convict someone. It might be enough to investigate someone, but there should be corroborating evidence.
When people say "fingerprints are bogus" they aren't saying that fingerprints literally don't exist, they are saying that the claims law enforcement/forensics people have made about their usefulness in uniquely identifying people are bogus. If there is enough error to discredit those claims, it's reasonable to say that the claims are bogus even if fingerprints do still have some value in identifying people.
We've had to warn you many times about breaking the site guidelines. There's no reason you couldn't have made your substantive point here without obvious flamebait like "feminists shriek". Moreover you've been continuing to break the rules in other places too:
I appreciate your positive contributions to HN, but if you keep this up we're going to end up having to ban you. I don't want to do that, so if you'd please review https://news.ycombinator.com/newsguidelines.html and properly fix this, that would be good.
The fact that we have a team of only 4 people doing this work when it should be a national effort with a couple billion in funding is a disgrace. Shout out to the Innocence Project. If you have the money, it's a great place if you want to make a change in our justice system.
It's a non-profit. The board doesn't do work. It's there to make sure you don't go around blowing the money or doing things contrary to the objective. Here it's presumably a way to get some cred and fundraising connections. They're probably net-positive in capital.
I'm setting up a non-profit and I need to find a board of independent directors. That's fine for me, but the thing will initially look like 1 worker, 5 directors. But it's not like I'm paying the 5.
Yeah I've never setup anything, but I assume a board more or less has a minimum size of several people that probably goes up the more complicated the setup is and as they get more involvement from outside groups. Unless they are paid positions, the size of the board really doesn't mean anything.
That would be the case if the organization wasn't mostly about what the Board and Advisory Board does - basically being a think tank/advisory board kind of deal.
And the stuff are just like secretarial services and such.
The Team page for that site has 115 people rather than 4. I'm sure it's a worthy cause and perhaps more people could help but there's no need to exaggerate...
"Kate Judson is a lawyer who often deals with crimes that did not occur. As the executive director of the Wisconsin-based Center for Integrity in Forensic Sciences (CIFS), her job is to examine ostensible scientific evidence to see whether it backs up prosecutors' claims.
"Some people who died were classified as victims of homicide when they were really the victim of illness, or accident, or suicide, or medical error—that kind of thing," says Judson. "We had a case of a family that lost their child. The caregiver was accused of attacking her. It was later discovered, based on new medical evidence, that the child had been really ill with a disease she was probably born with."
Evidence can't bring a child back, obviously. But it can get an innocent person out of jail. And it can give a grieving family some peace of mind. To learn that your child "was held and comforted in their last moments, instead of attacked," says Judson, "would be important to know."
When the center was founded four years ago, Judson left her job as a public defender to become its first employee. Now a staff of four works to keep bad science out of the courtroom. "
I concur. The Innocence Project and the CIFS are terrific organization. They are actively and effectively addressing flaws in a system that too often proves unjust to innocent citizens. They deserve all support and funding possible.
Until I was facing justice myself [1], I wrongly believed wrongful convictions were an extreme rarity. I actually discovered that innocent individuals are imprisoned every day, partly due to unreliable pseudo-scientific forensic methods [2]. I'm totally perplexed by society's apparent acceptance (or ignorance) of notable error rates in criminal convictions [3]. Why isn't it a major national cause already?
Nearly everything in forensic science has proven to be bullshit.
Drug dogs are worse than a coin flip - but the supreme court shrugged and said they're still valid.
Arson / "fire science" - nearly all of it had zero scientific backing and was bullshit.
Fingerprints aren't anywhere near as unique as we'd been told for decades, but they're still pushed, heavily.
"Lie detectors" are so useless they're illegal to use for employment or criminal matters.
Fiber/hair/handwriting analysis turns out to have also been bullshit.
All of it were just tools for police and prosecutors to secure a conviction for the person they decided was guilty, or to find someone a jury would think was guilty. Ie: poor and/or black.
You can add to your list the medical detection of child abuse from brain scans, aka abusive head trauma/shaken baby syndrome [1, 2]. Heavily unreliable without further evidence of physical trauma.
Do you have any sources? (Not calling BS, just curious.) I'm especially interested in drug-sniffing dogs, as I've heard that the dogs are often trained with specific commands to bark; then, when conducting a search, the officer will trigger the dog to bark with one of these commands, then use that bark as justification to search. In other words, the whole thing is just a prop. Not sure if there's truth to that, though; only what I've heard.
There are a bunch out there, this[0] is fairly representative of what I've read. What is worse, in my opinion, is the deference courts give to agencies using these dogs[1].
But, beyond that, surely you've met a few dogs in your life, no? One of the most obvious things about them is how badly they want to please their masters. What do you think the police want their dogs to do?
The scariest thing in the world is a jury of your peers. They don't know the law, even when explained to them. (I know this doesn't apply to everybody, but remember, juries are made up of all walks of life.) If an "expert" says almost anything, they believe it.
Now, if the experts are found to be lacking, it's even scarier.
Strong disagree. A judge can be dead wrong, but they have to put their opinion in writing and it's subject to appeal if there are fallacies in it. I would prefer a bench trial.
Yes! What I'm about to say is controversial, if it must be a jury pool, I'd rather have professional jurors then a jury of my peers. That way, how they voted would be on the record and the law books could be cited. (However we felt about the OJ trial circus, some of the jurors said they had already made up their minds even before the trial started!)
> An FBI analyst at his trial testified that there was just a one in 10 million chance that the hair found on a stocking mask at the crime scene belonged to someone other than Tribble.
Wasn’t before someone sent lizards DNA to one of those DNA sites and they got the results that it was human and Ashkenazi jew?
Yeah I believe it isn’t a solid proof as how the forensics folks make it sound.
One of those DNA sites aren't being used for forensic DNA analysis (estimating a probability that two DNA samples belong to the same person). Unlike 23 and me's claims of my ancestry, that kind of forensic analysis is scientifically validated and tested in court, including cases when the entire justice system is fighting tooth and nail to prevent a funny from being overturned.
Having done jury duty, I hate the concept of juries. Everyone is biased whether they like it or not and these biases converge pretty heavily, even amongst radically different people.
> Kate Judson is a lawyer who often deals with crimes that did not occur.
Oh, she deals with crimes alright, just not as charged in court. The crime is when the state uses bogus evidence to put an innocent person in prison. Or, you know, execute ("murder") an innocent person with made-up evidence[0].
I'll now wait patiently for the usual coterie of lawyers to comment about how shameful it is for non-lawyers to comment on the law, and how HN is full of dumb techies with terrible legal takes, and who, without exception or a shred of shame or self-awareness, post here on technical matters they have -- at best -- a fleeting grasp of.
It's very "late-stage capitalism" that we have all of this fantastic forensic technology but we aren't using it because we've outsourced evidence testing to the lowest fly-by-night bidder.
Even forensic DNA testing while potentially good science has ridiculous claims made about it - "one in a billion chance of being wrong". But it turns out they wrongly convict people using "unimpeachable" DNA evidence fairly often [1].
[1] https://www.science.org/content/article/forensics-gone-wrong...