Random stuff worth reading on the internet thread

[Pelado]

Crazy.

[Northwestcoug]

Another in the series of junk science in the courtroom. In this case, there's a former cop who teaches police departments around the country how to tell if someone who places a 911 call is lying or not. Along the way eager prosecutors and cops email how they can skirt the Daubert decision and get their findings admissible in court. The article does a good job showing how something that seems like just one piece of evidence becomes a lynchpin of conviction. It's a propublica article so it's long, but it's a good read:

https://www.propublica.org/article/9...ource=facebook

[Bo Diddley]

Noooo.....

It's not bad enough to have people coming up with faulty conclusions, now you want to trust computers running algorithms that nobody can clearly explain?

Only after AI demonstrated the ability to reliably pattern match, and that it can do it more reliably than humans.

[Pelado]

Let both sides present evidence and let the jury decide. Blanket ban of an entire class of evidence doesn’t seem rational to me.

I definitely did not expect this position from you. According to the article, individual ballistics characteristics are not at all scientific. They have presumably resulted in a lot of unwarranted convictions of innocent defendants.

That said, they've probably also allowed for the conviction of guilty defendants who might have otherwise gone free.

[Scott R Nelson]

Yeah, I totally agree with you after reading the article. I'd still like to see all the evidence presented at that hearing though. It would be interesting to see where AI could take the field of pattern matching.

Noooo.....

It's not bad enough to have people coming up with faulty conclusions, now you want to trust computers running algorithms that nobody can clearly explain?

[Bo Diddley]

This is the same attitude that let bite mark analysis fester in forensic cases for decades, and put innocent people in jail.

Yeah, I totally agree with you after reading the article. I'd still like to see all the evidence presented at that hearing though. It would be interesting to see where AI could take the field of pattern matching.

[Northwestcoug]

Let both sides present evidence and let the jury decide. Blanket ban of an entire class of evidence doesn’t seem rational to me.

This is the same attitude that let bite mark analysis fester in forensic cases for decades, and put innocent people in jail.

[Jeff Lebowski]

yes, the implication that all pattern analysis has some inherent bias. Humans tend to look for patterns naturally. Computer aided analysis could be better, but if the algorithms came from a human they can still be suspect.



did you read the article? In the best case scenarios, the false positive rate of calling two fired cartridges/bullets an exact match is ~20%. If you include ‘inconclusive’ results (remember, as a policy the FBI crime lab doesn’t exonerate any bullet fired; it is either called an ‘exact match’ or ‘inconclusive’), that rate shoots up to ~50%. A coin toss.

This ruling is exactly the thing that hopefully will generate more rational use of forensics in court cases.

Let both sides present evidence and let the jury decide. Blanket ban of an entire class of evidence doesn’t seem rational to me.

[Northwestcoug]

I took it to mean all sorts of pattern matching is suspect--not just ballistics.

yes, the implication that all pattern analysis has some inherent bias. Humans tend to look for patterns naturally. Computer aided analysis could be better, but if the algorithms came from a human they can still be suspect.

Seems like a dumb ruling.

did you read the article? In the best case scenarios, the false positive rate of calling two fired cartridges/bullets an exact match is ~20%. If you include ‘inconclusive’ results (remember, as a policy the FBI crime lab doesn’t exonerate any bullet fired; it is either called an ‘exact match’ or ‘inconclusive’), that rate shoots up to ~50%. A coin toss.

This ruling is exactly the thing that hopefully will generate more rational use of forensics in court cases.

[Bo Diddley]

Seems like a dumb ruling.

I'd like to see what evidence was presented.

[Pelado]

Seems like a dumb ruling.

How so?

[Jeff Lebowski]

Seems like a dumb ruling.

[Bo Diddley]

Since I don't own any firearms I'm guessing that I ought to be safe from having to worry about a wrongful conviction from bullet matching.

I took it to mean all sorts of pattern matching is suspect--not just ballistics.

[Scott R Nelson]

A Chicago circuit court judge barred the use of ballistics testing in a trial, which will probably reverberate nationally. This looong article explains the 'science' of ballistics firearm testing, how it is incorrectly viewed by both courtroom and lay people, and why in most cases it should be thrown out. At the very least if you read it, you will be convinced to lawyer up if there is ballistics evidence when you're a defendant in a trial:

https://radleybalko.substack.com/p/d...e-case-against

Okay, that was interesting.

I'm wondering if the defense could somehow test the person comparing two bullets to find out their error rate. I was thinking to gather 100 9mm pistols, maybe from the army or a large police department. Fire one bullet from each into some clay or something, then carefully retrieve it and stick it in a snack size ziplock bag. Do the same with the ejected cartridge. Put a 4-digit random number on each bag then note those in an Excel spreadsheet matching up with the gun serial number. To make it more interesting, don't save the bullets from about five of the tests and fire five other guns a second time to get the 100 bullets for testing. Then hand a box of bagged bullets, a box of bagged cartridges, and the pile of guns to the expert so see how many they can match. Include three boxes of the exact same ammuntion to simplify that part. That would produce some great percentage numbers for correct match, incorrect match, and inconclusive. I'm curious how long that would take the expert to do.

Since I don't own any firearms I'm guessing that I ought to be safe from having to worry about a wrongful conviction from bullet matching.

[Northwestcoug]

Thanks - very interesting read. Kind of shocking, really.

I’ve probably said this before, but Radley Balko is an excellent journalist. A sensible non-crazy libertarian-lite who has made police and judicial misconduct reporting his life’s work.