unsupported key premise is that the dental patterns are unique
October 20, 2022 9:31 AM Subscribe
Forensic Bitemark Analysis Not Supported by Sufficient Data, NIST Draft Review Finds.
"The National Institute of Standards and Technology (NIST) has reviewed the scientific foundations of bitemark analysis, a forensic technique in which marks on the skin of a biting victim are compared with the teeth of a suspected biter."
Forensic odontologists have known this since at least the 1990s when I was in dental school.
A person's occlusion isn't like a fingerprint and bite marks can pretty much only be used to exclude a suspect. This was noted in the link by the FO's own board.
posted by drstrangelove at 9:39 AM on October 20, 2022 [17 favorites]
A person's occlusion isn't like a fingerprint and bite marks can pretty much only be used to exclude a suspect. This was noted in the link by the FO's own board.
posted by drstrangelove at 9:39 AM on October 20, 2022 [17 favorites]
drstrangelove: thanks for that information! I didn't know that, but I always thought analyzing bite-marks seemed pretty suspect, like something made up for TV like "blood spatter analyst" on Dexter.
Then I found out that "blood spatter analyst" was already a real thing! And I wept for humankind.
I suspect that this will not lead to much change in how criminal investigations are conducted -- I mean, some cops still use psychics -- but one can hope.
posted by Saxon Kane at 10:00 AM on October 20, 2022
Then I found out that "blood spatter analyst" was already a real thing! And I wept for humankind.
I suspect that this will not lead to much change in how criminal investigations are conducted -- I mean, some cops still use psychics -- but one can hope.
posted by Saxon Kane at 10:00 AM on October 20, 2022
(Sort of) MeFi's own Radley Balko, who has a long history of writing about bite mark shenanigans including a terrific book, has a good-but-maddening take on the consequences (or utter lack thereof) that this is likely to have in prosecutorial decisionmaking on his new Substack.
posted by AgentRocket at 10:11 AM on October 20, 2022 [19 favorites]
posted by AgentRocket at 10:11 AM on October 20, 2022 [19 favorites]
Hasn't it been shown that pretty much all forensic "science", even fingerprinting, is basically astrology, except for DNA analysis and even then only under certain conditions?
posted by star gentle uterus at 10:48 AM on October 20, 2022 [16 favorites]
posted by star gentle uterus at 10:48 AM on October 20, 2022 [16 favorites]
A person's occlusion isn't like a fingerprint and bite marks can pretty much only be used to exclude a suspect. This was noted in the link by the FO's own board.
Reading the article it apparently goes further and says that the analysis can't even be used to exclude a subject, nor can forensic dentists even agree on patterns being human adult, child, animal or non-bite in some cases.
posted by BrotherCaine at 10:49 AM on October 20, 2022 [4 favorites]
Reading the article it apparently goes further and says that the analysis can't even be used to exclude a subject, nor can forensic dentists even agree on patterns being human adult, child, animal or non-bite in some cases.
posted by BrotherCaine at 10:49 AM on October 20, 2022 [4 favorites]
Hasn't it been shown that pretty much all forensic "science", even fingerprinting, is basically astrology, except for DNA analysis and even then only under certain conditions?
Pretty much, yeah.
posted by rhymedirective at 10:50 AM on October 20, 2022
Pretty much, yeah.
posted by rhymedirective at 10:50 AM on October 20, 2022
Forensic science is mostly make-believe, ranging from innocent error to systemic fraud.
posted by Pope Guilty at 10:51 AM on October 20, 2022 [11 favorites]
posted by Pope Guilty at 10:51 AM on October 20, 2022 [11 favorites]
The excellent podcast Criminal touched on this topic in June, in the episode The Sailor's Teeth.
> In 1982, forensic dentists examined the teeth of thousands of sailors stationed on an aircraft carrier called the USS Carl Vinson in Newport News, Virginia. It’s been called “the largest dental dragnet likely in U.S. history.”
posted by cyranix at 10:58 AM on October 20, 2022 [6 favorites]
> In 1982, forensic dentists examined the teeth of thousands of sailors stationed on an aircraft carrier called the USS Carl Vinson in Newport News, Virginia. It’s been called “the largest dental dragnet likely in U.S. history.”
posted by cyranix at 10:58 AM on October 20, 2022 [6 favorites]
I was racking my brain trying to think of where I already knew this from - thank you cyranix. Love the Criminal podcast!
posted by widdershins at 11:55 AM on October 20, 2022
posted by widdershins at 11:55 AM on October 20, 2022
Forensic science is inherently broken because it starts from proving an assumption rather than looking at what "evidence" points towards, nothing is double blinded and you get paid for results that match what the cops want.
posted by Ferreous at 11:57 AM on October 20, 2022 [22 favorites]
posted by Ferreous at 11:57 AM on October 20, 2022 [22 favorites]
Not surprisingly, one more thing that is much more about making the "evidence" fit the narrative that prosecutors want to tell rather than establishing actual truths.
posted by BigHeartedGuy at 1:20 PM on October 20, 2022
posted by BigHeartedGuy at 1:20 PM on October 20, 2022
except for DNA analysis
My understanding is that DNA is 1 in a billion, so you can match six or seven people in the world.
posted by rhizome at 1:51 PM on October 20, 2022 [1 favorite]
My understanding is that DNA is 1 in a billion, so you can match six or seven people in the world.
posted by rhizome at 1:51 PM on October 20, 2022 [1 favorite]
the actual work most forensic scientists (and not hired gun "expert witnesses") do day to day is spending 95% of their time testing drugs to prove they're drugs. Doing the same analytic tests over and over to say that yes, this white powder is chemically confirmed as cocaine, this unknown plant matter is confirmed as marijuana etc etc. Cops aren't out there solving shit unless it's someone important or the media makes them look bad. Drug busts on the other hand are easy and make crime numbers go up.
posted by Ferreous at 2:20 PM on October 20, 2022 [2 favorites]
posted by Ferreous at 2:20 PM on October 20, 2022 [2 favorites]
The landmark study cited and linked in TFA is 279 pages long. It recommended this review of bitemark analysis 13 yrs ago and it just got done, when, in truth 4 out of 5 dentists surveyed would have said: yep, that's BS.
posted by OHenryPacey at 2:28 PM on October 20, 2022 [4 favorites]
posted by OHenryPacey at 2:28 PM on October 20, 2022 [4 favorites]
Police authorities are addicted to forensics which will allow them to convict whoever they want to convict badly enough.
Something new will arise as enough courts and defense attorneys get wise to bite mark analysis; it would be nice to be able to guess what it might be and preempt it.
posted by jamjam at 2:29 PM on October 20, 2022
Something new will arise as enough courts and defense attorneys get wise to bite mark analysis; it would be nice to be able to guess what it might be and preempt it.
posted by jamjam at 2:29 PM on October 20, 2022
Same thing with hair analysis
To add insult, after the report showing hair analysis was bogus, the government closed down the commission that created the report:
"In 2017, new Attorney General Jeff Sessions, appointed by President Donald Trump, announced that this investigation would be suspended, at the same time that he announced the end of a forensic science commission that had been working to establish standards on several tests and to improve accuracy; it was a "partnership with independent scientists to raise forensic science standards".[18] Independent scientists, prosecutors, defense counsel and judges criticized ending the commission, saying that the criminal justice system needed to rely on the best science."
posted by soylent00FF00 at 3:27 PM on October 20, 2022 [6 favorites]
To add insult, after the report showing hair analysis was bogus, the government closed down the commission that created the report:
"In 2017, new Attorney General Jeff Sessions, appointed by President Donald Trump, announced that this investigation would be suspended, at the same time that he announced the end of a forensic science commission that had been working to establish standards on several tests and to improve accuracy; it was a "partnership with independent scientists to raise forensic science standards".[18] Independent scientists, prosecutors, defense counsel and judges criticized ending the commission, saying that the criminal justice system needed to rely on the best science."
posted by soylent00FF00 at 3:27 PM on October 20, 2022 [6 favorites]
the actual work most forensic scientists (and not hired gun "expert witnesses") do day to day is spending 95% of their time testing drugs to prove they're drugs. Doing the same analytic tests over and over to say that yes, this white powder is chemically confirmed as cocaine, this unknown plant matter is confirmed as marijuana etc etc. Cops aren't out there solving shit unless it's someone important or the media makes them look bad. Drug busts on the other hand are easy and make crime numbers go up.
And sometimes the drug tests are just thousands of lies.
Same thing with hair analysis
Finding out that basically every kind of "fiber analysis" is basically just looking at fibers under a microscope and deciding if they look the same or not, to the point that forensic scientists have proven themselves unable to distinguish human hair, dog hair, and even thread, was a big "oh it's all a scam, isn't it?" moment for me.
posted by Pope Guilty at 3:33 PM on October 20, 2022 [6 favorites]
And sometimes the drug tests are just thousands of lies.
Same thing with hair analysis
Finding out that basically every kind of "fiber analysis" is basically just looking at fibers under a microscope and deciding if they look the same or not, to the point that forensic scientists have proven themselves unable to distinguish human hair, dog hair, and even thread, was a big "oh it's all a scam, isn't it?" moment for me.
posted by Pope Guilty at 3:33 PM on October 20, 2022 [6 favorites]
this unknown plant matter
I know you may not have meant it as a double-entendre, but that's a big part of the problem, too. They spend so much time confirming that the substance is a drug, but they don't actually establish a chain of custody verified by independent sources to show that the suspect was in possession.
posted by explosion at 3:49 PM on October 20, 2022 [1 favorite]
I know you may not have meant it as a double-entendre, but that's a big part of the problem, too. They spend so much time confirming that the substance is a drug, but they don't actually establish a chain of custody verified by independent sources to show that the suspect was in possession.
posted by explosion at 3:49 PM on October 20, 2022 [1 favorite]
This is the only forensic expert I'd trust.
posted by signal at 4:16 PM on October 20, 2022 [2 favorites]
posted by signal at 4:16 PM on October 20, 2022 [2 favorites]
When my kid was maybe a year and a half or so old, another kid bit him on the face and left quite a bite-batwing bruise. Kids explore with their mouths at that age, it wasn't even a tussle. The child care staff saw the kid do it, so this is just a thing I felt compelled to share and has no bearing at all.
posted by theora55 at 4:35 PM on October 20, 2022 [2 favorites]
posted by theora55 at 4:35 PM on October 20, 2022 [2 favorites]
Hasn't it been shown that pretty much all forensic "science", even fingerprinting, is basically astrology, except for DNA analysis and even then only under certain conditions?
Fingerprints are complicated. As a means of identification, comparison between two well recorded sets of fingerprints (or even a single fingerprint) is pretty good. And modern digital techniques have well characterized error rates.
Latent fingerprint analysis, on the other hand, in which partial fingerprints are "lifted" from surfaces to identify a putative suspect, is very, very sketchy.
posted by mr_roboto at 5:38 PM on October 20, 2022 [6 favorites]
Fingerprints are complicated. As a means of identification, comparison between two well recorded sets of fingerprints (or even a single fingerprint) is pretty good. And modern digital techniques have well characterized error rates.
Latent fingerprint analysis, on the other hand, in which partial fingerprints are "lifted" from surfaces to identify a putative suspect, is very, very sketchy.
posted by mr_roboto at 5:38 PM on October 20, 2022 [6 favorites]
On a whole other set of hands...
Automated License Plate Readers (ALPRS)
Body-Worn Cameras
Cell-site Simulators/ISMI Catchers
Drones/Unmanned Aerial Vehicles
Surveillance Cameras
Face Recognition
Tattoo Recognition
Iris Recognition
and Electronic Monitoring...
to open a case or two of canned worms are all categories in the Electronic Fontiers Foundation's
STREET LEVEL SURVEILLANCE.
Make what you will of that.
posted by y2karl at 6:29 PM on October 20, 2022 [3 favorites]
Automated License Plate Readers (ALPRS)
Body-Worn Cameras
Cell-site Simulators/ISMI Catchers
Drones/Unmanned Aerial Vehicles
Surveillance Cameras
Face Recognition
Tattoo Recognition
Iris Recognition
and Electronic Monitoring...
to open a case or two of canned worms are all categories in the Electronic Fontiers Foundation's
STREET LEVEL SURVEILLANCE.
Make what you will of that.
posted by y2karl at 6:29 PM on October 20, 2022 [3 favorites]
Bruce the Shark was framed. "Bite radius," my red ass.
posted by delfin at 6:59 PM on October 20, 2022
posted by delfin at 6:59 PM on October 20, 2022
Add to the list fake arson investigations, which resulted in the execution of an man who was almost certainly innocent.
posted by Mr.Know-it-some at 6:35 AM on October 21, 2022 [5 favorites]
posted by Mr.Know-it-some at 6:35 AM on October 21, 2022 [5 favorites]
In the same way, canine scent detection is complicated: accuracy is good if dog and handler are regularly training in double blind scenarios that accurately mimic working conditions and incorrect alerts are unrewarded/dogs willing to make them are removed from programs.
On the other hand, it is extremely easy to deliberately or accidentally condition dogs to alert to handler expectations rather than actually alerting to odor, and even easier to teach a dog to "alert" to a very subtle cue from a handler whether or not the odor of interest is present. Mysteriously this is much more of a problem for drug detection K9 units than USDA Customs beagles or cadaver detection teams. Unfortunately people tend to conceptualize dogs as wholly honest and infallible, which also complicates jury response.
posted by sciatrix at 7:11 AM on October 21, 2022 [4 favorites]
On the other hand, it is extremely easy to deliberately or accidentally condition dogs to alert to handler expectations rather than actually alerting to odor, and even easier to teach a dog to "alert" to a very subtle cue from a handler whether or not the odor of interest is present. Mysteriously this is much more of a problem for drug detection K9 units than USDA Customs beagles or cadaver detection teams. Unfortunately people tend to conceptualize dogs as wholly honest and infallible, which also complicates jury response.
posted by sciatrix at 7:11 AM on October 21, 2022 [4 favorites]
Fingerprints are complicated. As a means of identification, comparison between two well recorded sets of fingerprints (or even a single fingerprint) is pretty good. And modern digital techniques have well characterized error rates.Any sources for that, mr_roboto? I was recently reading about this on Wikipedia and it seemed to say the opposite:
Furthermore, academics have argued that the error rate in matching fingerprints has not been adequately studied. And it has been argued that fingerprint evidence has no secure statistical foundation.posted by Cogito at 9:38 AM on October 21, 2022
> Hasn't it been shown that pretty much all forensic "science", even fingerprinting, is basically astrology, except for DNA analysis and even then only under certain conditions?
Yep! Everything, basically, from arson detection to fingerprints to cadaver dogs to blood spatter analysis, fiber analysis, handwriting analysis, it's all fake. Basically you pick a suspect, take some evidence, and find an expert who will connect your suspect to the evidence and hope that the person doesn't have the money to hire a counter-expert who is convincing enough to throw doubt on the evidence. Forensics in TV shows have broken the brains of the jury pool enough to believe it all to be true, so the burden of proof comes to rest on the accused yelling and screaming that no, that isn't one of my hairs, that's A DOGS HAIR. Even DNA analysis has so many pitfalls: if the labwork isn't perfect, you can get a positive match when there shouldn't be.
posted by dis_integration at 9:42 AM on October 21, 2022 [2 favorites]
Yep! Everything, basically, from arson detection to fingerprints to cadaver dogs to blood spatter analysis, fiber analysis, handwriting analysis, it's all fake. Basically you pick a suspect, take some evidence, and find an expert who will connect your suspect to the evidence and hope that the person doesn't have the money to hire a counter-expert who is convincing enough to throw doubt on the evidence. Forensics in TV shows have broken the brains of the jury pool enough to believe it all to be true, so the burden of proof comes to rest on the accused yelling and screaming that no, that isn't one of my hairs, that's A DOGS HAIR. Even DNA analysis has so many pitfalls: if the labwork isn't perfect, you can get a positive match when there shouldn't be.
posted by dis_integration at 9:42 AM on October 21, 2022 [2 favorites]
One of the other issues is that the daubert standard uses judges as "gatekeepers" for admissibility of evidence but it turns out most judges aren't actually good or qualified to judge whether evidence is scientifically sound. They're just as susceptible to being flimflammed by some expert witness who has "testified in hundreds of cases" in a fancy suit using complete nonsense slides to posit a connection between utterly unconnected pieces of information.
posted by Ferreous at 10:36 AM on October 21, 2022 [1 favorite]
posted by Ferreous at 10:36 AM on October 21, 2022 [1 favorite]
Or not even being flimflammed but simply trusting people they view as being in the same economic/social class as them over eyewitnesses or the accused.
posted by Ferreous at 10:43 AM on October 21, 2022 [3 favorites]
posted by Ferreous at 10:43 AM on October 21, 2022 [3 favorites]
Latent fingerprint analysis, on the other hand, in which partial fingerprints are "lifted" from surfaces to identify a putative suspect, is very, very sketchy.
My understanding is that, even given that the "official" odds of a full-print match are make-believe, a total ass-pull, prosecutors have a tendency to present fragmentary prints as definitive matches and cite the (again, make-believe) odds of a match as though they were presenting full, detailed prints obtained under laboratory conditions.
posted by Pope Guilty at 11:36 AM on October 21, 2022 [3 favorites]
My understanding is that, even given that the "official" odds of a full-print match are make-believe, a total ass-pull, prosecutors have a tendency to present fragmentary prints as definitive matches and cite the (again, make-believe) odds of a match as though they were presenting full, detailed prints obtained under laboratory conditions.
posted by Pope Guilty at 11:36 AM on October 21, 2022 [3 favorites]
Any sources for that, mr_roboto? I was recently reading about this on Wikipedia and it seemed to say the opposite:
The big gap in statistical validation is in the latent prints, partly because they use non-standardized subjective methods. I'm sure we could always use more data for use in controlled environments, but there are lots of studies on that. If they were like bite marks, you couldn't use them to do things like unlock digital devices.
This is not to defend current practices in the criminal justice system. I still remember, about a decade ago, researchers sent a pair of prints to 15 labs, informing them ahead of time that they were from a famous case where they had been incorrectly matched. Fourteen of the labs found them to be different, one said they were the same. The kicker, of course, was the researchers lied--they sent each lab a pair that they the lab itself had previously said under oath were the same. It was a great example of how vulnerable to priming the techniques are.
Forensic science is inherently broken because it starts from proving an assumption rather than looking at what "evidence" points towards, nothing is double blinded and you get paid for results that match what the cops want.
Sort of a nitpick, but double blind is a buzzword that doesn't really mean anything when it comes to testing inanimate object--it's basically for human research, and means both the scientist and the subject don't know the treatment method.
Blinding (without the "double") should be used in forensics for sure--though it is used more often these days. My anecdote directly above is a good example why.
But it's not like it's a standard prerequisite for good science with validated methods. I spent 15 years in a lab as an analytical chemist--in a "real" scientific field--I don't think I ran a single blinded sample. (TBF there's an argument blinding should be used more often in science, but validated reproducible methods often make that essentially a moot point. You trust readings from a scale or a measuring cup without "blinding.")
posted by mark k at 12:15 PM on October 21, 2022 [7 favorites]
Furthermore, academics have argued that the error rate in matching fingerprints has not been adequately studied. And it has been argued that fingerprint evidence has no secure statistical foundation.Note that this is referring to fingerprints as evidence--ie, the latent partial prints that mr_roboto said were sketchy. For valid uses they were talking about something like your DMV print matching your security clearance background print as having utility.
The big gap in statistical validation is in the latent prints, partly because they use non-standardized subjective methods. I'm sure we could always use more data for use in controlled environments, but there are lots of studies on that. If they were like bite marks, you couldn't use them to do things like unlock digital devices.
This is not to defend current practices in the criminal justice system. I still remember, about a decade ago, researchers sent a pair of prints to 15 labs, informing them ahead of time that they were from a famous case where they had been incorrectly matched. Fourteen of the labs found them to be different, one said they were the same. The kicker, of course, was the researchers lied--they sent each lab a pair that they the lab itself had previously said under oath were the same. It was a great example of how vulnerable to priming the techniques are.
Forensic science is inherently broken because it starts from proving an assumption rather than looking at what "evidence" points towards, nothing is double blinded and you get paid for results that match what the cops want.
Sort of a nitpick, but double blind is a buzzword that doesn't really mean anything when it comes to testing inanimate object--it's basically for human research, and means both the scientist and the subject don't know the treatment method.
Blinding (without the "double") should be used in forensics for sure--though it is used more often these days. My anecdote directly above is a good example why.
But it's not like it's a standard prerequisite for good science with validated methods. I spent 15 years in a lab as an analytical chemist--in a "real" scientific field--I don't think I ran a single blinded sample. (TBF there's an argument blinding should be used more often in science, but validated reproducible methods often make that essentially a moot point. You trust readings from a scale or a measuring cup without "blinding.")
posted by mark k at 12:15 PM on October 21, 2022 [7 favorites]
Any sources for that, mr_roboto? I was recently reading about this on Wikipedia and it seemed to say the opposite:
NIST has been studying this for awhile. Here's a summary of a recent study on contactless readers.
Again, this is fingerprinting technology for identity verification that might be used for border entry control or device or room access. The exemplar fingerprints were taken using a standard technique and they are compared to a fingerprint presented in a way meant to maximize accuracy.
This is very, very different from latent fingerprint analysis as used in criminal forensics.
posted by mr_roboto at 12:20 PM on October 21, 2022 [2 favorites]
NIST has been studying this for awhile. Here's a summary of a recent study on contactless readers.
Again, this is fingerprinting technology for identity verification that might be used for border entry control or device or room access. The exemplar fingerprints were taken using a standard technique and they are compared to a fingerprint presented in a way meant to maximize accuracy.
This is very, very different from latent fingerprint analysis as used in criminal forensics.
posted by mr_roboto at 12:20 PM on October 21, 2022 [2 favorites]
Sort of a nitpick, but double blind is a buzzword that doesn't really mean anything when it comes to testing inanimate object--it's basically for human research, and means both the scientist and the subject don't know the treatment method.
In the context of scentwork--since I used the term--it generally refers to exercises in which neither dog nor any human handler present know the location of the planted detection target before the dog is released to search.
posted by sciatrix at 10:38 PM on October 21, 2022
In the context of scentwork--since I used the term--it generally refers to exercises in which neither dog nor any human handler present know the location of the planted detection target before the dog is released to search.
posted by sciatrix at 10:38 PM on October 21, 2022
Soon to be followed (i hope) by body language behaviour analysis.
posted by TWinbrook8 at 2:07 AM on October 22, 2022 [1 favorite]
posted by TWinbrook8 at 2:07 AM on October 22, 2022 [1 favorite]
« Older Corporate Influence on Inflation | the most frequently told story about ed-tech Newer »
This thread has been archived and is closed to new comments
posted by Saxon Kane at 9:36 AM on October 20, 2022 [12 favorites]