Myth of Fingerprints?

SLAM: debunk creationism, pseudoscience, and superstitions. Discuss logic and morality.

Moderator: Alyrium Denryle

Post Reply
User avatar
Elfdart
The Anti-Shep
Posts: 10692
Joined: 2004-04-28 11:32pm

Myth of Fingerprints?

Post by Elfdart »

This article on Counterpunch is very interesting.

http://counterpunch.org/cockburn02202006.html

It makes the case that in many instances, the use of fingerprint evidence, like phrenology, polygraphs and other "scientific" evidence from the bad old days is flim-flam.
Few law enforcement institutions have been so thoroughly discredited in recent years as the FBI's forensic lab. In 1997 the Bureau's inspector general of the time issued a devastating report, stigmatizing one instance after another of mishandled and contaminated evidence, inept technicians, and outright fabrication. The IG concluded that there were "serious and credible allegations of incompetence" and perjured courtroom testimony.

CounterPunch's view is that taken as a whole, forensic evidence as used by prosecutors is inherently untrustworthy. For example, for years many people went to prison on the basis of the claims of a North Carolina anthropologist, Louise Robbins. She helped send people to prison or to Death Row with her self-proclaimed power to identify criminals through shoe prints. As an excellent recent Chicago Tribune series on forensic humbug recalled, on occasion she even said she could use the method to determine a person's height, sex and race. Robbins died in 1987, her memory compromised by the conclusion of many Appeals Courts that her methodology was bosh. There have been similarly hollow claims for lip prints and ear prints, all of (added "of") them invoked by their supporters as "100 per cent reliable" and believed by juries too easily impressed by passionate invocations to 100 per cent reliable scientific data.

Of course the apex forensic hero of prosecutors, long promoted as the bottom line in reliability--at least until the arrival of DNA matching--has been the fingerprint.

Fingerprints entered the arsenal of police and prosecutors in the late nineteenth century, touted as "scientific" in the manner of other fashionable methods of that time in the identification of supposed criminals, such as phrenology. A prime salesman was Francis Galton, Charles Darwin's cousin and a founding huckster for the bogus "science " of eugenics. Actually fingerprints, at least in modern times, found their original use in the efforts of a British colonial administrator to intimidate his Indian laborers (whose faces he could not distinguish) from turning up more than once to get paid. He'd make a great show of scrutinizing the fingerprints he insisted they daub on his ledger book. Then, as now, the use of the so-called "unique fingerprint" has been histrionic , not scientific. In 1995, so the Chicago Tribune series discovered, " one of the only independent proficiency tests of fingerprint examiners in U.S. crime labs found that nearly a quarter reported false positives, meaning they declared prints identical even though they were not--the sort of mistakes that can lead to wrongful convictions or arrests."

Decade after decade people have been sent to prison for years or dispatched to the death cell, solely on the basis of a single, even a partial print. So great is the resonance of the phrase "a perfect match" that defense lawyers throw in the towel, as judge and jury listen to the assured conclusions of the FBI's analysts who virtually monopolize the fingerprint industry in the U.S.A. Overseas, in London's Scotland Yard for example, the same mesmerizing "certainty" held sway, and still does. In the U.S.A., part of the mystique stems from the "one discrepancy rule" which has supposedly governed the FBI's fingerprint analysis. The rule says that identifications are subject to a standard of "100 per cent certainty" where a single difference in appearance is supposed to preclude identification.

The 1997 lab scandals threw a shadow over the FBI's forensic procedures as a whole and the criminal defense bar began to raise protests against prosecutorial use of latent fingerprint identification evidence, as produced by FBI procedures. In 2002 Judge Louis Pollak, in a case in Pennsylvania, initially ruled that the FBI's fingerprint matching criteria fell below new standards of forensic reliability (the Daubert standards) stipulated by the U.S. Supreme Court. Ultimately he was persuaded that the FBI's fingerprint lab had never made a mistake. In 2004, in U.S. v. Mitchell, the Third Circuit Court of Appeals upheld these same procedures.

Now at last, in 2006, the FBI's current inspector general, Glenn Fine, has grudgingly administered what should properly be regarded as the deathblow to fingerprint evidence as used by the FBI and indeed by law enforcement generally.

The case reviewed by Inspector General Fine, at the request of U.S. Rep John Conyers and U.S. Senator Russell Feingold, concerns the false arrest by the FBI of Brandon Mayfield, a lawyer from Beaverton, Oregon.
Mayfield was accused of taking part in the Madrid bombings based on a single print, supposedly from Mayfield's index finger. The Spanish authorities arrested an Algerian named Ouhnane Daoud on the grounds that the print was supposedly from his middle finger. Luckily the judge in Seattle dismissed the case, though not before Mayfield was arrested, perp-walked in front of the media, thrown in jail and roughed up behind bars.

Does anyone else find it suspicious that the "evidence" just happened to finger (no pun intended) Mayfield?
Through covert surveillance they had learned that Mayfield was married to an Egyptian woman, had recently converted to Islam, was a regular attendee at the Bailal mosque in Portland, and had as one of his clients in a child custody dispute an American Muslim called Jeffrey Battle. Battle, a black man, had just been convicted of trying to go to Afghanistan to fight for the Taliban.
This case looks like a frame-up to me.
Image
User avatar
Lord Zentei
Space Elf Psyker
Posts: 8742
Joined: 2004-11-22 02:49am
Location: Ulthwé Craftworld, plotting the downfall of the Imperium.

Post by Lord Zentei »

Doubtless fingerprinting is an imprecise science, and doubtless there are instances where it is being misused if partial prints are deemed enough for a conviction.

However, I would hesitate to dismiss it entirely: modern pattern recognition programs might give the discipline the rigour it has lacked hitherto.

Through covert surveillance they had learned that Mayfield was married to an Egyptian woman, had recently converted to Islam, was a regular attendee at the Bailal mosque in Portland, and had as one of his clients in a child custody dispute an American Muslim called Jeffrey Battle. Battle, a black man, had just been convicted of trying to go to Afghanistan to fight for the Taliban.
This case looks like a frame-up to me.
Indeed. I might point out that this in and of itself does not incriminate fingerprinting per se: even with the most rigorous forensic science, prosecuters can still provide bullshit evidence if the judiciary fails in its vigilance.
CotK <mew> | HAB | JL | MM | TTC | Cybertron

TAX THE CHURCHES! - Lord Zentei TTC Supreme Grand Prophet

And the LORD said, Let there be Bosons! Yea and let there be Bosoms too!
I'd rather be the great great grandson of a demon ninja than some jackass who grew potatos. -- Covenant
Dead cows don't fart. -- CJvR
...and I like strudel! :mrgreen: -- Asuka
User avatar
Surlethe
HATES GRADING
Posts: 12267
Joined: 2004-12-29 03:41pm

Post by Surlethe »

It seems to me like the article is using the incompetence of some crime labs to impreach fingerprinting as a whole. If they want to do so, it's going to take more than accusations of misconduct on the part of crime labs to overturn objective science.
A Government founded upon justice, and recognizing the equal rights of all men; claiming higher authority for existence, or sanction for its laws, that nature, reason, and the regularly ascertained will of the people; steadily refusing to put its sword and purse in the service of any religious creed or family is a standing offense to most of the Governments of the world, and to some narrow and bigoted people among ourselves.
F. Douglass
User avatar
Elfdart
The Anti-Shep
Posts: 10692
Joined: 2004-04-28 11:32pm

Post by Elfdart »

This part might explain why:
The prints were given to Kenneth R. Moses of San Francisco, an SFPD veteran who runs a company called Forensic Identification Services which, among other things, proclaims its skills in "computer enhancement of fingerprints". It was "quite difficult", Moses said, because of "blurring and some blotting out", but yes, the FBI had it right, and there was "100 per cent certainty" that one of the prints on the blue bag in Madrid derived from the left index finger of Brandon Mayfield.

Moses transmitted this confident opinion by phone to Judge Jones on the morning of May 19. Immediately following Moses' assertion, the U.S. attorney stepped forward to confide to Judge Jones dismaying news from Madrid from the Spanish police that very morning. The news "cast some doubt on the identification". This information, he added, "was classified or potentially classified".
So it's not just mischief at the FBI lab. This private lab also decided it was Mayfield's print. This smells like the 60 Minutes sting on polygraphs where results just happened to fit preconceived notions. The front company told the polygraphers that they thought a certain employee had stolen from the company. Of course nobody stole anything, but the polygraph operators insisted there was no doubt the guy embezzled.

This looks like the same thing.
Image
User avatar
Durandal
Bile-Driven Hate Machine
Posts: 17927
Joined: 2002-07-03 06:26pm
Location: Silicon Valley, CA
Contact:

Post by Durandal »

This sounds like typical defense lawyer bullshit, to be honest. The basic premise that each print is unique is sound. But individual analysts use their own criteria for determining what is and isn't a match. One might declare a 6-point match to be a match while another would not. It boils down to human judgment. This article casts doubt on one specific case, but certainly not the whole of criminal forensics.

Polygraphs are a different story. Their fundamental premise is that lying causes changes in vital signs, and there isn't a shred of evidence to back that up. Polygraphing is generally used as a breakdown technique anyway. The interrogator uses it as a tool to intimidate the person being questioned. Having the subject think you're omniscient is a good way to break him down.

When a government agency polygraphs you for an employment position or something, they'll usually take you into a room and prep you. They implant an idea in your head that a certain behavior (like lying to authority figures) is considered extremely bad. They just assume that everyone, at some point, has lied to an authority figure in their lives. But they prep you to lie to that kind of question by making you think telling the truth will automatically fail you. That's how they get a reading on you.

It's a technique, no different from good-cop-bad-cop. Sometimes it works, sometimes it doesn't.
Damien Sorresso

"Ever see what them computa bitchez do to numbas? It ain't natural. Numbas ain't supposed to be code, they supposed to quantify shit."
- The Onion
User avatar
Oddysseus
Padawan Learner
Posts: 415
Joined: 2003-06-28 01:12am
Location: Operating secretly in the heartland of the Homeland.

Post by Oddysseus »

Elfdart wrote:This part might explain why:
The prints were given to Kenneth R. Moses of San Francisco, an SFPD veteran who runs a company called Forensic Identification Services which, among other things, proclaims its skills in "computer enhancement of fingerprints". It was "quite difficult", Moses said, because of "blurring and some blotting out", but yes, the FBI had it right, and there was "100 per cent certainty" that one of the prints on the blue bag in Madrid derived from the left index finger of Brandon Mayfield.

Moses transmitted this confident opinion by phone to Judge Jones on the morning of May 19. Immediately following Moses' assertion, the U.S. attorney stepped forward to confide to Judge Jones dismaying news from Madrid from the Spanish police that very morning. The news "cast some doubt on the identification". This information, he added, "was classified or potentially classified".
So it's not just mischief at the FBI lab. This private lab also decided it was Mayfield's print. This smells like the 60 Minutes sting on polygraphs where results just happened to fit preconceived notions. The front company told the polygraphers that they thought a certain employee had stolen from the company. Of course nobody stole anything, but the polygraph operators insisted there was no doubt the guy embezzled.

This looks like the same thing.
When I first heard about this case, I thought it was passed around and studied inside the FBI, garnering support for the conclusion, then went to the specialist. But they may have been exaggerating.


Looking at the case, I see that they got what the guy’s lawyer called "an image of substandard quality". The FBI fed it into their supercomputer for matching fingerprints, the Automated Fingerprint Identification System. Then three FBI examiners, Terry Green, John T. Massey, and Michael Wieners, separately, narrowed it down to Mayfield. Moses was court appointed and backed up the FBI position. Then the Spanish said they were wrong and that the prints pointed to an Algerian. Then the FBI sent people to Spain to look at the data and conceded the Spanish were right.

And I love the FBI claim, "FBI had told [the defense attorney] the error was the first ever by the agency's fingerprint analysis group." Is that even probable? Statistically speaking.
- Odd Jack, Jaded Skeptic
--- jadedskeptic.blogspot.com
- "The meme for blind faith secures its own perpetuation by the simple unconscious expedient of discouraging rational inquiry."
"The universe is a strange and wondrous place. The truth is quite odd enough to need no help from pseudoscientific charlatans." - Richard Dawkins
User avatar
Jalinth
Jedi Council Member
Posts: 1577
Joined: 2004-01-09 05:51pm
Location: The Wet coast of Canada

Post by Jalinth »

Durandal wrote:This sounds like typical defense lawyer bullshit, to be honest. The basic premise that each print is unique is sound. But individual analysts use their own criteria for determining what is and isn't a match. One might declare a 6-point match to be a match while another would not. It boils down to human judgment. This article casts doubt on one specific case, but certainly not the whole of criminal forensics.
Your own comments basically condemn fingerprinting. "individual analysts use their own criteria for determining what is and isn't a match". Unless you have some objective standards or criteria for matching, you wander into the "art" category rather than the science. Is this a Picasso or a forgery of Picasso.

Too much subjectivity is lethal. At least fingerprinters need to be able to quantify that X number of points matching or higher is an acceptable (for court) match with an error rate of no more than Y.
User avatar
Master of Ossus
Darkest Knight
Posts: 18213
Joined: 2002-07-11 01:35am
Location: California

Post by Master of Ossus »

Jalinth wrote:Your own comments basically condemn fingerprinting. "individual analysts use their own criteria for determining what is and isn't a match". Unless you have some objective standards or criteria for matching, you wander into the "art" category rather than the science. Is this a Picasso or a forgery of Picasso.

Too much subjectivity is lethal. At least fingerprinters need to be able to quantify that X number of points matching or higher is an acceptable (for court) match with an error rate of no more than Y.
:roll:

Different statisticians use different cut-offs for different things. That doesn't invalidate statistics. The fingerprinters almost certainly differ only in their threshold for identification.
"Sometimes I think you WANT us to fail." "Shut up, just shut up!" -Two Guys from Kabul

Latinum Star Recipient; Hacker's Cross Award Winner

"one soler flar can vapririze the planit or malt the nickl in lass than millasacit" -Bagara1000

"Happiness is just a Flaming Moe away."
User avatar
Wyrm
Jedi Council Member
Posts: 2206
Joined: 2005-09-02 01:10pm
Location: In the sand, pooping hallucinogenic goodness.

Post by Wyrm »

Jalinth wrote:
Durandal wrote:This sounds like typical defense lawyer bullshit, to be honest. The basic premise that each print is unique is sound. But individual analysts use their own criteria for determining what is and isn't a match. One might declare a 6-point match to be a match while another would not. It boils down to human judgment. This article casts doubt on one specific case, but certainly not the whole of criminal forensics.
Your own comments basically condemn fingerprinting. "individual analysts use their own criteria for determining what is and isn't a match". Unless you have some objective standards or criteria for matching, you wander into the "art" category rather than the science. Is this a Picasso or a forgery of Picasso.

Too much subjectivity is lethal. At least fingerprinters need to be able to quantify that X number of points matching or higher is an acceptable (for court) match with an error rate of no more than Y.
More critically, I've seen no good assessment of the false positive and false negative rates of fingerprinting. Such a determination is VITAL for assessing its usefulness as a forensic tool. Let's take a concrete example to examine why:

Suppose, for the sake of argument that NYC has 10 million people, one of whom has committed a crime. Suppose further that the only piece of evidence you have is a small sample of DNA from the perp, and it matches the suspect you have in coustody. Suppose, finally, that if the suspect is the perpetrator, then the DNA test will definitely match, but if the suspect is not the purpetrator, then the probability that the test would match is 1 in one million. What are the odds that the suspect is the perpetrator?

A naïve answer would lead you to conclude that the odds that the suspect is the purpetrator is a million to one. This is the usual interpretation used by prosecutors... but its wrong. Suppose you tested ALL 10 million people in NYC against your DNA evidence. Then, 10 of the innocent NYers will test positive by chance alone (the perpetrator, of course, will test positive). Therefore, the true odds that the suspect is the perpetrator, based on this single piece of evidence alone, is 1:10, not 1e6:1. That's a huge difference, and this particular error in reasoning (assuming P(match|innocent) = P(innocent|match)) is called the prosecutor's fallacy.

Note that we were able to make this assessment because we know the false positive rate of the DNA test (such tests are by nature statistical tests, given the fact that the entire genome is not compared, only select parts of them). Also, note that this is based on the DNA sampling being a single piece of evidence. In practice, the test is applied to a small pool of likely individuals, and it is the combination of the DNA evidence and the evidence that establishes the pool that really settles the person's guilt. Even if each piece of evidence is actually pretty weak, the convergence of many pieces evidence is usually robust.

But we don't have the false positive rates of fingerprinting in general. Therefore, fingerprinting's usefulness as a forensic tool cannot yet be measured. Is it a positive match weak piece of evidence, or is it a strong piece of evidence? We just don't know.

I would therefore never convict a fellow on fingerprint evidence alone. I would need some corroborating evidence, enough to widdle the field down to a few likely suspects. I would demand no less from a DNA match, and it is almost definitely a stronger test.

Going back to Elfdart's point, the reasons for singling out Mayfield seem weak at best. What was the evidence that one of the perps was even from the US? What evidence points to the perpetrator being married to a muslim woman (as opposed being to a muslim man; ie, why wasn't the perp a woman), a recent convertee (as opposed to someone raised muslim), and had child coustody issues with a muslim who was convicted of trying to join the Taliban? I don't see how they can piece that profile together from any evidence they could find in the bomb debris.
Darth Wong on Strollers vs. Assholes: "There were days when I wished that my stroller had weapons on it."
wilfulton on Bible genetics: "If two screaming lunatics copulate in front of another screaming lunatic, the result will be yet another screaming lunatic. 8)"
SirNitram: "The nation of France is a theory, not a fact. It should therefore be approached with an open mind, and critically debated and considered."

Cornivore! | BAN-WATCH CANE: XVII | WWJDFAKB? - What Would Jesus Do... For a Klondike Bar? | Evil Bayesian Conspiracy
User avatar
Master of Ossus
Darkest Knight
Posts: 18213
Joined: 2002-07-11 01:35am
Location: California

Post by Master of Ossus »

Wyrm wrote:But we don't have the false positive rates of fingerprinting in general. Therefore, fingerprinting's usefulness as a forensic tool cannot yet be measured. Is it a positive match weak piece of evidence, or is it a strong piece of evidence? We just don't know.
Bullshit. Just because you haven't taken the time to do a three minute Google search does not mean that the body of scientific knowledge has skipped over your concerns. Do you really think that a normal member of the population has all of the evidence the scientific community has to analyze data just because they're concerned about the results?

Do you really believe that forensic science would completely ignore this ludicrously basic analysis when attempting to evaluate a form of possible evidence?
I would therefore never convict a fellow on fingerprint evidence alone. I would need some corroborating evidence, enough to widdle the field down to a few likely suspects. I would demand no less from a DNA match, and it is almost definitely a stronger test.
You are a fucking idiot and a moral coward. People like you allow thousands of felons to go free every year because of your unrealistic expectations of the justice system.
The so-called "50K study" took a set of 50,000 pre-existing images of fingerprints and compared each one electronically against the whole of the data set, producing a grand total of 2.5 billion comparisons. It concluded that the chances of each image being mistaken for any of the other 49,999 images were vanishingly small, at 1 in 10^97.
One unpublished study may go some way to answering the critics. It documents the results of exercises in which 92 students with at least one year's training had to match archive and mock "crime scene" prints. Only two out of 5861 of these comparisons were incorrect, an error rate of 0.034 per cent. Kasey Wertheim, a private consultant who co-authored the study, told New Scientist that the results have been submitted for publication.
Even biometric devices using comparatively low-quality tests and algorithms available in the public sphere boast of false positive and false negative rates of less than 1%.

In conclusion, thanks for fucking up our criminal justice system, your ignorant piece of shit.

Edit: Copy-paste didn't carry the exponent, properly.
Last edited by Master of Ossus on 2006-02-21 03:25pm, edited 1 time in total.
"Sometimes I think you WANT us to fail." "Shut up, just shut up!" -Two Guys from Kabul

Latinum Star Recipient; Hacker's Cross Award Winner

"one soler flar can vapririze the planit or malt the nickl in lass than millasacit" -Bagara1000

"Happiness is just a Flaming Moe away."
User avatar
drachefly
Jedi Master
Posts: 1323
Joined: 2004-10-13 12:24pm

Post by drachefly »

Master of Ossus wrote:Different statisticians use different cut-offs for different things. That doesn't invalidate statistics. The fingerprinters almost certainly differ only in their threshold for identification.
And half-way decent statisticians don't say "100% sure".
User avatar
Master of Ossus
Darkest Knight
Posts: 18213
Joined: 2002-07-11 01:35am
Location: California

Post by Master of Ossus »

drachefly wrote:And half-way decent statisticians don't say "100% sure".
What's your point? I'm sure some people are unqualified to do fingerprint analyses, or misinterpret their results colloquially. That does not discredit fingerprinting, in general, which I have shown to be quite accurate--theoretically rivalling DNA evidence and practically only a cut below that.
"Sometimes I think you WANT us to fail." "Shut up, just shut up!" -Two Guys from Kabul

Latinum Star Recipient; Hacker's Cross Award Winner

"one soler flar can vapririze the planit or malt the nickl in lass than millasacit" -Bagara1000

"Happiness is just a Flaming Moe away."
User avatar
drachefly
Jedi Master
Posts: 1323
Joined: 2004-10-13 12:24pm

Post by drachefly »

What I'm getting at is, if a fingerprinter says "100% sure these are the same person" they are delusional or lying.
Take lab 1. They use normal standards. They report 100% certainty.
Now take lab 2. They use very stringent standards. Are they totally wasting their time by doing the extra steps? No. Lab 1 was wrong to reort 100% certainty. But Lab 2 reports 100% certainty.
Now take lab 3...

Even in the absence of subversion of the evidence, 100% is unattainable.
User avatar
Master of Ossus
Darkest Knight
Posts: 18213
Joined: 2002-07-11 01:35am
Location: California

Post by Master of Ossus »

drachefly wrote:What I'm getting at is, if a fingerprinter says "100% sure these are the same person" they are delusional or lying.
Take lab 1. They use normal standards. They report 100% certainty.
Now take lab 2. They use very stringent standards. Are they totally wasting their time by doing the extra steps? No. Lab 1 was wrong to reort 100% certainty. But Lab 2 reports 100% certainty.
Now take lab 3...

Even in the absence of subversion of the evidence, 100% is unattainable.
For all intents and purposes, though, 10^-97=0. I guess there's technically reason for doubt, there, but not reasonable doubt. It is not realistic to demand such a high standard of evidence, although the technician involved clearly misrepresented the lab's results.
"Sometimes I think you WANT us to fail." "Shut up, just shut up!" -Two Guys from Kabul

Latinum Star Recipient; Hacker's Cross Award Winner

"one soler flar can vapririze the planit or malt the nickl in lass than millasacit" -Bagara1000

"Happiness is just a Flaming Moe away."
User avatar
drachefly
Jedi Master
Posts: 1323
Joined: 2004-10-13 12:24pm

Post by drachefly »

Yes, 10^-97 is effectively zero... but these crime labs are calling 10^-7 or so zero -- how many errors do you think they've made? How many cases do you think they've handled?
Do you think these numbers match their claimed 100% reliability?


In the more corrupt cases, they're calling 1 zero.
User avatar
Wyrm
Jedi Council Member
Posts: 2206
Joined: 2005-09-02 01:10pm
Location: In the sand, pooping hallucinogenic goodness.

Post by Wyrm »

Master of Ossus wrote:
Wyrm wrote:But we don't have the false positive rates of fingerprinting in general. Therefore, fingerprinting's usefulness as a forensic tool cannot yet be measured. Is it a positive match weak piece of evidence, or is it a strong piece of evidence? We just don't know.
Bullshit. Just because you haven't taken the time to do a three minute Google search does not mean that the body of scientific knowledge has skipped over your concerns. Do you really think that a normal member of the population has all of the evidence the scientific community has to analyze data just because they're concerned about the results?

Do you really believe that forensic science would completely ignore this ludicrously basic analysis when attempting to evaluate a form of possible evidence?
Okay, I deserved that. :oops:

I'd like to point out some things, though: the article states that this is for computerized matching. If I'm going to ascribe a confidence level of 99% to a match, I'd have to know this. What's the false positive rate for traditional Mk I eye matching? It would be improper to award traditional fingerprint matching the accuracy of computerized matching.

Furthermore, there's mention of "most accurate systems", implying varying accuracy. If I'm ever at trial, I'd like to know the system and its performance characteristics.

False negative rate? If the defense were to present evidence of a non-match, I'd like to know how much weight to give it.

Furthermore, the accuracy of 99.6% for the best system was for a two finger match, and the more fingers, the more accurate the match. This implies, of course, the fewer fingers, the less accurate the match. What's the performance for a single full finger? A partial print?

No, don't bother Googling; knowing that this stuff has been examined has put me a little at ease. However, my point should be clear. The accuracy of fingerprinting is not 99%; it's contitional, on a great many things.
The so-called "50K study" took a set of 50,000 pre-existing images of fingerprints and compared each one electronically against the whole of the data set, producing a grand total of 2.5 billion comparisons. It concluded that the chances of each image being mistaken for any of the other 49,999 images were vanishingly small, at 1 in 10^97.
I call foul on that. This is comparing a set of static snapshots of fingerprints against each other, not comparing them with a fingerprint collected in the field from a finger known to be in the database. Take any two arbitrary images off your hard drive. Chances are, they will not match up. This is what it sounds like to me.
One unpublished study may go some way to answering the critics. It documents the results of exercises in which 92 students with at least one year's training had to match archive and mock "crime scene" prints. Only two out of 5861 of these comparisons were incorrect, an error rate of 0.034 per cent. Kasey Wertheim, a private consultant who co-authored the study, told New Scientist that the results have been submitted for publication.
How big was the archive? How was the "crime scene" prints made? Did they take the print out of the archive and pass it around the students, or was the finger that made the original archive print used to make a new "crime scene" print (possibly degraded)? If so, how good was the "crime scene" print? Was it pristine (like taking the original fingerprint), or did the mock "perpetrator" touch the object in a natural manner, the print lifted, and sent off for comparison? I can EASILY see how these various conditions can affect the results, so I ask, were they addressed?
I would therefore never convict a fellow on fingerprint evidence alone. I would need some corroborating evidence, enough to widdle the field down to a few likely suspects. I would demand no less from a DNA match, and it is almost definitely a stronger test.
You are a fucking idiot and a moral coward. People like you allow thousands of felons to go free every year because of your unrealistic expectations of the justice system.
Ah, now this is a different matter entirely.

IF the procescution can convince me that the police have widdled down the pool of likely perpetrators to, say, on the order of 10,000 persons, and the suspect is a part of this pool before applying the test to that pool, THEN this additional DNA evidence is pretty damning, and I'm likely to convict. The pool may be implied rather than explicit, but there must be some form of prior evidence that this accused person is part of a restricted pool of about this size. If the pool is much larger than this, I'd start demanding an additional piece of evidence from the procecution. Yes, even the vaunted DNA match.

Does this mean I'd never convict on a piece of DNA evidence? No. It's good evidence, but it can't pick the one unique guilty person out of a LARGE group of people. There has to be other evidence besides it (even evidence that the prosecution may not even realize is evidence). If the prior odds of guilt was 1:1 before introducing a DNA match, then I would convict that sucker... the posterior odds of guilt would indeed be 1e6:1.

Also, even at 99% true positive rate, I'd consider a pool size of 101 people much too large for me to consider convicting on a fingerprint alone (assuming that the match probability really is 99%, see above), because the posterior odds are 1:1, which does not fit my definition of reasonable doubt, and I doubt yours does either. The tipping point would probably come around a pool size of 11 people, but don't quote me on that.
Even biometric devices using comparatively low-quality tests and algorithms available in the public sphere boast of false positive and false negative rates of less than 1%.
No, not if the biometric is the only evidence the prosecution brings forth, and even in combination with a pool of 100 possible perpetrators, because the posterior odds of guilt is still only 1:1.
In conclusion, thanks for fucking up our criminal justice system, your ignorant piece of shit.
Fuck you, Ossus. I know how to apply conditional probabilities, I understand the prosecutor's fallacy (and the defense attourney's fallacy), and I do not believe a 1:1 posterior odds of guilt constitutes "guilty beyond a reasonable doubt." I don't think you would either.

I think the real problem lies elsewhere.
Darth Wong on Strollers vs. Assholes: "There were days when I wished that my stroller had weapons on it."
wilfulton on Bible genetics: "If two screaming lunatics copulate in front of another screaming lunatic, the result will be yet another screaming lunatic. 8)"
SirNitram: "The nation of France is a theory, not a fact. It should therefore be approached with an open mind, and critically debated and considered."

Cornivore! | BAN-WATCH CANE: XVII | WWJDFAKB? - What Would Jesus Do... For a Klondike Bar? | Evil Bayesian Conspiracy
User avatar
Durandal
Bile-Driven Hate Machine
Posts: 17927
Joined: 2002-07-03 06:26pm
Location: Silicon Valley, CA
Contact:

Post by Durandal »

Jalinth wrote:Your own comments basically condemn fingerprinting. "individual analysts use their own criteria for determining what is and isn't a match". Unless you have some objective standards or criteria for matching, you wander into the "art" category rather than the science. Is this a Picasso or a forgery of Picasso.
What a load of shit. Fingerprinting is not simply holding up two transparencies and seeing if they line up. The objective standard is pattern recognition and point matching. Statisticians routinely declare a certain value to be acceptable error. Does that mean the entirety of statistics is nothing more than an art?
Too much subjectivity is lethal. At least fingerprinters need to be able to quantify that X number of points matching or higher is an acceptable (for court) match with an error rate of no more than Y.
There is no set standard for matching. The whole point is that if an examiner identifies a 6-point match, but visually, he can tell that the prints are the same, he can declare a match with the same certainty as a 10-point match. Objectivity is maintained, since anyone can look at the prints themselves and determine whether the examiner was flawed.
Damien Sorresso

"Ever see what them computa bitchez do to numbas? It ain't natural. Numbas ain't supposed to be code, they supposed to quantify shit."
- The Onion
User avatar
drachefly
Jedi Master
Posts: 1323
Joined: 2004-10-13 12:24pm

Post by drachefly »

If anyone can do it, why are there experts?

What then is the difference between a 6 or 10 point match?

... and so forth.


by the way, Wyrm, it should be noted that fingerprints may be able to more strongly exclude some people than others. So, a pool of 101 people would be fine if only one of them has the general shape of the fingerprint... heck, a pool of 10,000 would be okay if somehow you managed to get none of the right shape. Of course, that's really unlikely for such a large group, but you get the idea.
User avatar
Wyrm
Jedi Council Member
Posts: 2206
Joined: 2005-09-02 01:10pm
Location: In the sand, pooping hallucinogenic goodness.

Post by Wyrm »

drachefly wrote:by the way, Wyrm, it should be noted that fingerprints may be able to more strongly exclude some people than others. So, a pool of 101 people would be fine if only one of them has the general shape of the fingerprint... heck, a pool of 10,000 would be okay if somehow you managed to get none of the right shape. Of course, that's really unlikely for such a large group, but you get the idea.
Then you understand part of what I'm getting at: the accuracy of a fingerprint match is not 99%, but conditional.

Certainly, if I were on a jury, and the expert witness testifying on some match was able to put forth some convincing explanation that the match is especially good, then I will take that into account. The caveat here is that the expert has to make a convincing argument in that direction. On the other hand, if the expert makes it clear (or I begin to strongly suspect) that the match is especially bad, then it will be given less weight accordingly.

The type of fingerprint database searches like you see on CSI, where you plug in a thumbprint and out pops a single name, is a misuse of the forensics. Even with the best computer matching systems availible with a full four-finger match, which has a false positive rate of 0.01% (thanx, Ossus), then searching from the one unique perpetrator in the IAFIS database (47 million records) would expect to yield ~4700 matching records, only one of which is the perpetrator's. Any random person from that group has a 99.97% chance of being innocent.

What's happening here is, although the probability of the crime scene print matching an individual innocent person is slim, there are many, many opportunities to match innocent people to that print. Therefore, the innocent "noise" drowns out the guilty "signal." This problem only gets worse as the database gets bigger. Indeed, if our criterion for an initial match is to return the best (say) ten matching candidates, the false positive rate would have to be no more than 10/N (where N is the size of the database). In short, it has to be Fucking Awesome!

The upshot of this is that the fingerprint match does not tell you that Mr/Ms. X is guilty, or even that the prints were made by the same finger, but that the two prints fall within some variance criterion (and for good matching paradigms, a metric related to the reliability of fit). It is then up to the forensics experts to properly put that piece of evidence into context with the body of evidence.

Even good forensics are destructive when misused. To use fingerprints in a blind search for a single individual is a misuse. It can easily poison an investigation focusing myoptically on the first schmuck who turns up in the search. Fingerprint forensics are best saved until you have a short list of likely perpetrators (for an appropriate definition of "short").
Darth Wong on Strollers vs. Assholes: "There were days when I wished that my stroller had weapons on it."
wilfulton on Bible genetics: "If two screaming lunatics copulate in front of another screaming lunatic, the result will be yet another screaming lunatic. 8)"
SirNitram: "The nation of France is a theory, not a fact. It should therefore be approached with an open mind, and critically debated and considered."

Cornivore! | BAN-WATCH CANE: XVII | WWJDFAKB? - What Would Jesus Do... For a Klondike Bar? | Evil Bayesian Conspiracy
User avatar
drachefly
Jedi Master
Posts: 1323
Joined: 2004-10-13 12:24pm

Post by drachefly »

I was not trying to correct you, merely add upon what you were saying. We are in complete agreement.
Post Reply