Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

FBI testimony on hair analysis contained errors in 90% of cases (2015)

fbi.gov

Post reply on HN
Print view View original

71–77 of 77 posts

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#71

Post by c4ptnjack »

thaumasiotes wrote: post #59
slavboj wrote: post #53

Earlier quoted context omitted.

See, your job is to get accurate results because clients are paying to know if they have a problem or not. The FBI's job is to get convictions, so their lab's job is to create railroadable pseudo-evidence.

Mold lawsuits are very common; the money in mold testing is to always find mold. People who honestly want to know whether they have a problem are dwarfed in number by people who honestly want a certificate that says mold was found.

Working at an industrial hygiene company that works with large train companies in particular, doing studies on everything from asbestos to air particulates.

The entire industry is around finding where those issues exist, using accurate data to track it, and solve whatever problems come up to avoid the large risk of getting sued by employees or communities nearby.

I did QA on all the data that came in from labs and matched it all the various input samples we sent off. Incredibly thorough testing procedures, redundancy, and tracking over time was the name of the game for my companies entire business. They've been successful for 30+ years doing that too.

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#72

Post by ars »

jacobolus wrote: post #14
mkl95 wrote: post #9

Within two weeks I have learnt that bite mark and hair analysis are full of BS. Is there some list of all these shoddy methods?

Polygraphs, fingerprints, DNA, face matching, forensic ballistics, forensic psychology, witness testimony (especially police testimony), ...

> witness testimony

There are two kinds of witness testimony - the most common is where a witness sees a stranger and testifies about it. Those have little credibility.

But there's another kind, where the witness actually knows the person, those are far more credible. Historically only this type was permitted, but over time things seem to have shifted to the first type.

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#73

Post by kwhitefoot »

Natsu wrote: post #20
kwhitefoot wrote: post #10

It doesn't seem that they will take action to prevent failures of a similar general kind occurring again. There should be legislation regarding the quality of forensic analysis of all kinds.

> It doesn't seem that they will take action to prevent failures of a similar general kind occurring again. That seems to contradict this part of the article: > The Department has been working together with the Innocence Project and NACDL to address errors made in statements by FBI examiners regarding microscopic hair analysis in the context of testimony and laboratory reports. Such statements are no longer being mad…

That's taking action to prevent errors of the same kind. My, poorly expressed, point was that all forensic analysis needs to be held to a defined and agreed quality standard. And both quality assurance and quality control should be applied to such things to ensure that the standard is met.

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#74

Post by nibbleshifter »

MerelyMortal wrote: post #33
jacobolus wrote: post #14

Earlier quoted context omitted.

Polygraphs, fingerprints, DNA, face matching, forensic ballistics, forensic psychology, witness testimony (especially police testimony), ...

I'm willing to believe that fingerprints are unique until proven otherwise, though I would need to see the actual finger prints and not some computer report claiming they "match". I also know that fingerprints can be faked.

Manual fingerprint comparison is also deeply unreliable tbh

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#76

Post by MerelyMortal »

throwaway74829 wrote: post #41
MerelyMortal wrote: post #33

Earlier quoted context omitted.

I'm willing to believe that fingerprints are unique until proven otherwise, though I would need to see the actual finger prints and not some computer report claiming they "match". I also know that fingerprints can be faked.

Fingerprints are unique in the technical sense (i.e. no one has the exact same set of fingerprints) -- but matching is a completely different issue. I don't know how far forensic science has come since a decade ago when I learned some of it, but at that point they were running "feature matching" instead of full-fingerprint matching. In other words, they looked for aberrancies (e.g. "islands," "deltas," "bifurcations"…

I think this is the case you're refering to: https://www.theregister.com/2004/05/26/fbi_madrid_blunder/

Re: FBI testimony on hair analysis contained errors in 90% of cases (2015)

#77

Post by hedora »

akira2501 wrote: post #54
hedora wrote: post #23

So, if this is used in a trial where I am on the jury, I'll assume it implies a 90% chance the evidence says the defendant is innocent, plus a 5 to 9.9% chance of a false positive with the underlying test. Got it. Moving on. Edit: I was being snarky but, on reflection, if they are falling back on hair analysis, it implies they have a piece of the guilty party's hair and the defendant's hair, but are not willing to pr…

> I'll assume it implies a 90% chance the evidence says the defendant is innocent Evidence isn't there to _prove_ anything, else we wouldn't need to ask anyone to be a juror. That's why we use "preponderance of evidence" and "beyond reasonable doubt" as our metric. It also assumes that all juries, once they see bite mark analysis, presume it's magic then simply decide they're going to convict at that moment and shut…

The use of this test, which requires exactly the same forensic evidence as a decisive one says a lot about the prosecution. The only reason to use this is if you think a more accurate (and cheaper) test would prove the suspect's innocence.

That sounds like a crazy conspiracy, but, from the headline 90% of the time, the test is misadministered. That suggests that, indeed, all DA's that use this tactic are "corrupt to the point of no longer caring about justice and truth".

Do you have an alternate explanation for a 90% error rate?

Post reply on HN