IAI resolution 2010-18 (The Detail #462, 26/07/10)

[Iain McKie]

Thanks for responding Michele.

My comments are not aimed at any lack of support for Shirley from the IAI. From the time I was invited to address its conference in Miami its members have kept that support going and the findings of the ‘Y7’ Committee’ came at an extremely important time.

I am referring to the following postings by ’16 points’.

by 16_Points » Mon Aug 16, 2010 9:42 am

‘………..I've seen non-matching prints with 14, 15, even 16 minutiae that agree very well with no visible disagreements. These are rare occurrences, but they do happen. I have not seen the SCRO prints. As I put in the parenthetical, claiming 20 minutiae in agreement is only meaningful if the examiner's observations are accurate.’

by 16_Points » Mon Aug 16, 2010 3:29 pm

‘……..I knew that it wasn't an identification because the questioned print was actually an experimental control of known origin. Hence we know with 100% certainty that it is an exclusion. The exemplar was obtained via an AFIS search optimized to produce a close non-matching candidate (the actual AFIS search was conducted by a distinguished colleague who showed me the results at an IAI lecture).

This is the difference between the knowledge gained via casework and the knowledge gained via controlled experimentation. Furthermore, it would be very interesting to see the likelihood ratio generated by a 14 point close non-match vis-a-vis other 14 point matches. Everyone would learn something useful.’

My understanding is that ‘16 Points’ has claimed to have seen two fingerprints, ‘with 14, 15, even 16 minutiae that agree very well with no visible disagreements.’, that originate from different donors.

’16 Points’ claims to have seen one such result at an IAI lecture.

My assertion is that if ’16 Points’ has got his/her facts straight then such findings undermine the very basis on which fingerprint evidence is accepted by the courts.

Basically I am asking that the IAI/experts have the accuracy and relevance of this alleged ‘close non-match' assessed.

My own personal view would be that there is some logical explanation for ’16 Points’ assertions that will not bring the science down but surely such a potentially damaging claim must be laid to rest.

[Michele]

I see. You think it’s necessary for practitioners, organizations, and the discipline to respond to the allegations of 16 points even though he has no proof to back it up.

‘IF’ he produces some substance, ‘THEN’ I think people would be more willing to address the concern. For now, his allegation is just a hypothetical situation which seems meritless to me.

[Iain McKie]

Thanks Michele,

I always found that the best way of ensuring ‘facts’ were ‘facts’ was to keep testing them whenever the opportunity arose.

I also find it strange whenever people challenge a ‘fact’ there are those who immediately label them as critics (in the negative sense).

My hope in becoming involved in this debate was that ’16 Points’ would come back with further explanation or proof that could be objectively assessed. I assume I cannot be criticised for this.

I would however agree that until '16 Points' or someone else does respond there is little that can be done and also agree that those who make claims should support them or at least assist others in their assessment.

I would also accept that I have been influenced by the ‘head in the sand’ approach to fingerprint mistakes over the years in the UK and immediately grow suspicious when this appears to be happening.

I like your postings 'end quote',

“The best way to escape from a problem is to solve it."- Alan Saporta

To this might I add,

“You won’t find a solution by saying there is no problem” – William Rotsler

Iain

[Outsider]

I have always been sceptical of the claim that a close non-match that would fool an experienced examiner has never been found. Would this claim not be more correctly stated as “no non-match has ever been found that would fool an experienced examiner when the examiner knows in advance that it is a non-match”?

To test the claim properly a very large number of close match tests would have to be made under normal casework conditions. Examiners would have to think that the test exemplar is from a potential suspect or someone on the elimination list.

The most important test would be to check the integrity of verification. Verifiers would have to carry out their work thinking that the test pair has already been declared identical by a respected colleague or by a highly respected fingerprint department. The verifier would have to think that disagreeing with the original decision would lead to a great deal of trouble possibly putting careers in jeopardy, including his own.

I agree with Iain McKie that 100% certainty is not possible in any human endeavour. A pure mathematical principle like “2 and 2 equals 4” is not a human endeavour, neither is the rising and setting of the sun. I think Sandra Wiese equating a fingerprint ID to a “fact” involves a leap of faith of the sort that Professor Champod describes. It may be a fact that two or more experts reach the same conclusion but this does not make it a fact that the identified person depositied the mark.

Let’s not forget that forensic experts are not “finders of fact”. That is the job of police officers, officials who decide to prosecute, and juries. It is one thing for two or more fingerprint experts (who are fallible human beings) to be certain that an identified person deposited a mark but it is another thing for the police and legal officials to be certain that an identified person is lying just because they dispute an ID (as happened to Shirley McKie and, possibly, Brandon Mayfield).

[mark beck]

Outsider.... "A pure mathematical principle like 2 and 2 equals 4 is not a human endeavour..." True enough, but neither is the "Fact" that a friction ridge impression is a result of a touch of a unique and singular source. Any and every impression of friction ridge skin can only be from the exact portion of the person's skin that made the touch that left the impression. Whether or not the impression can be attributed to the source depends on many factors, but the fact that one source made the impression is unquestionable. Can a person erroneously or carelessly or absent mindedly add two and two and get five....maybe. Can a latent print examiner make a similar type error in making a latent print comparison...it has happened on rare occassion. Leaving a fingerprint is not a "human endeavour" ... it is physics ... Locard's third... with maybe a little chemistry thrown in. It is unfortunate that there is so much confusion as to what is fact and what is philosophy and what is just plain BS. Luckily we have persons such as Sandra willing to step up to defend the principles of the science from spurious claims such as those exampled by 16 pts. Her clear and concise explanations deserve serious regard and not the semantical tap dance of your response.

[16_Points]

I see things warmed up a little bit in my absence. Let me say this, whether or not you find a 15 point close non-match credible is irrelevant. A 14, 15, or 16 point close non-match does not threaten or weaken latent print identification. If anything, it provides a valuable opportunity to refine fingerprint identification methodology.

The point, again, is to put this within the context of probability models. Some previous posters expressed doubts about the utility of likelihood ratio models. I submit that these models are very useful. If we claim, as latent print examiners, that an identification can be made when the chance (i.e. probability) that another person could have made the crime scene impression is so small that it can be ignored, then we should be concerned with our ability to determine that probability. The very fact that the notion of a 15 point close non-match (with no discernible differences) provokes such skepticism among fingerprint experts suggests that this scenario has a low probability. You might even say the probability is so small that it can be ignored . . . it's a Black Swan.

To draw a parallel, the chance of dying in a plane crash is so small that you would have to travel by air everyday for several lifetimes to be at real risk of dying in a plane crash. Nonetheless, planes still crash, people die every year, and yet air travel is still extremely safe. The point being that the appearance of the Black Swan does not automatically invalidate all of our operating assumptions. However, if we make an error in estimating these probabilities, then we could have a problem. Setting up a likelihood ratio or other probabilistic model for fingerprints can only help us explore this possibility.

[kevin]

I thought everyone was calling the close non-match a doppleganger and not a black swan?

[Boyd Baumgartner]

Let me say this, whether or not you find a 15 point close non-match credible is irrelevant

I'd say to the Jury, the Statistician, the Defense Lawyer and a segment of the Examiners on this board it is very much relevant. The more pertinent question would be whether or not the person claiming it exists is credible.

You might even say the probability is so small that it can be ignored . . . it's a Black Swan

point being that the appearance of the Black Swan does not automatically invalidate all of our operating assumptions. However, if we make an error in estimating these probabilities, then we could have a problem. Setting up a likelihood ratio or other probabilistic model for fingerprints can only help us explore this possibility.

Don't these notions miss the point of black swan phenomenon? Black Swans are defined as having non calculable probabilities and are subjective functions of the observer. Your airplane example is a prime illustration of this. The chance of dying for someone who died as the result of a plane crash are 100%. As NNT stated, the black swan for the turkey who's been hand fed and raised by humans up until the day before Thanksgiving is not the same as the black swan for the butcher. The point is to not be the turkey. Statistics leave out the qualitative elements of the examination and are therefore impotent to close the door of vulnerability from a quantitative/qualitative standpoint.

I think the point here is more simply represented by the somewhat existentially less exciting notion of 'seeing is believing". If it exists, let's see it.

[mark beck]

That would not be a quote from me....in fact as I understand the concept, Black Swans are high impact events that occur which have an unpredictably low probability of occurance. As such, the 15 point non-match claimed by 16 pts is not a Black Swan ... because it didn't happen and it doesn't exist. As he said....it's "a notion"....in this case a fantasy.

[16_Points]

I'm going to have to disagree with Boyd. It's like saying that the risk of complications from LASIK eye surgery is less than 1%. But if you are that person, then your risk is 100%. Bravo. Well reasoned and totally useless in practice. Tell that to someone assessing the risk of surgery and see how much that helps the decision making process. It's meaningless unless you can define a set of parameters so that one can determine if he is in the high risk pool (the <1% that will have complications).

The purpose of the Black Swan . . . as in the book . . . was that Nassim Taleb made boat loads of money betting on Black Swan events. Why? Because the probability of these events were so low that the contracts (options etc.) were very cheap, pennies in fact. They were so cheap that he could make hundreds of ridiculous bets, and the pay outs were so large that he only had to hit one or two to make money. Simply put, rare events may be considered to be "practically impossible" but they do occasionally happen. But just because it happens, we don't overturn the whole decision making process. I thought I had made this point at least a couple of times (in response to McKie). Hence my statement that it's irrelevant whether or not you find a 15 point close non-match credible. Why? Because it doesn't change anything. So it doesn't matter if it's true (which it is) or whether I fabricated the scenario after going on a bender of malt liquor and peyote. Nothing about the fundamental nature of fingerprint identification changes (unless you can demonstrate that these types of close non-matches are common occurrences). That is the important conclusion.

Mark, yes, high impact events so on etc. Though I use it simply to mean very rare events which might be considered "impossible". If you consider a 15 point close non-match to be a fantasy, so be it. But I am amused by the responses of "I'll believe it when I see it." It's as if we're talking about unicorns and leprechauns as opposed to low probability events. What can I say? It's not my data. I only saw it. But if everyone's interest at this point is only to say, "It's a load of BS unless you show it to me." Then let's just move on.

[Taggart]

16 points,
I wonder if I could offer a few observations for your consideration. Given you are coming under some serious criticism by other posters you previously stated:

(the actual AFIS search was conducted by a distinguished colleague who showed me the results at an IAI lecture).

Why not name this ‘distinguished colleague’ who originally showed you this amazing result at the IAI lecture? I am imagining many experts may well know the individual and will be able to confirm your story. Take the pressure of yourself.
You also stated:

The exemplar was obtained via an AFIS search optimized to produce a close non-matching candidate

And that you believed

it was a beta test to see if such close non-matches were lurking in the AFIS database. Nonetheless, it's an interesting result.

I have no problem with these statements but when I asked if the images could be published you stated:

I wish the prints were available for publication. If my recollection is correct, it was outside the scope of the AFIS search agreement. I mean that my colleague did not have permission to publish any images. It would be nice if someone had the opportunity to do this as a study for publication.

I am struggling with this.
Why carry out any such experiments/tests if the results could never be published?
And can you are anyone else confirm exactly what the AFIS ‘agreement’ is over experiments/tests being carried out on the AFIS and in particular to the publication of any significant findings?
Given your colleague took the prints along to an IAI lecture would it have been outside any AFIS ‘agreement ‘for them to have handed their findings over to the IAI, given the massive significance of such a finding?
If your colleague was unable to ‘publish’ their findings, was there anything within the AFIS ‘agreements’ which allowed your colleague to ‘show’ the images to others, as they clearly were able to do with yourself?
As long as the donor is kept anonymous is there a problem? Given there is a claim here of a print being found with apparently 16 points in close agreement with none in disagreement, is this finding not so significant to the science that it needs to be shared?
You also stated in your recent posting that:

The very fact that the notion of a 15 point close non-match (with no discernible differences) provokes such skepticism among fingerprint experts suggests that this scenario has a low probability. You might even say the probability is so small that it can be ignored . . . it's a Black Swan.

I believe the vast majority of Experts would agree with this. That if asked they would state the possibility of such an occurrence would be so low that it could be ignored. However both you and your distinguished colleague are exceptions to the rule. If I understand you correctly both you and your colleague have both seen such an occurrence, a 15 point close non-match, with no discernible differences. Would you not admit that from your experience that your probability of such a phenomenon occurring has to be higher that others who have never seen such a print?
Out of interest how many others were shown these prints at the IAI Lecture you both attended? Was anyone in authority made aware of this amazing non-match? I find it hard to believe that your colleague only showed the prints to you and you alone.

[16_Points]

Struggle with it all you like. I maintain my position. I did not raise the point to generate a side discussion on the credibility of of a 15 point close non-match. Nor do I see why there is such a fuss. Do latent print examiners really feel threatened by such a claim? (Rhetorical question) For I cannot explain otherwise the keen interest. Unless it's simply a burning desire to impeach my integrity. I find it even more curious considering there were such vehement posts by examiners claiming that they do not count minutiae. If all these examiners have abandoned point counting, why does the number suddenly hold so much interest for people? (Rhetorical question) Furthermore, the question isn't why hasn't this been published? The question is why hasn't anyone else tried this before? (Also Rhetorical)

That's it. In summary, I'd like to reiterate my central argument which is that a statistical model based on minutiae is very useful and can only support and enhance latent print identifications.

[Taggart]

I guess I’ll just need to keep struggling without a single answer to any of my questions. Perhaps someone else can help me out with the AFIS ‘agreement’ about not publishing results of any experiments/testing carried out on the system.

I am unfortunately struggling even more with your latest posting.

I did not raise the point to generate a side discussion on the credibility of of a 15 point close non-match. Nor do I see why there is such a fuss.

Every year in British Courts individuals are being convicted on the basis of fingerprint evidence. Given there is a non-numeric standard here that could be anywhere from perhaps 10 points in agreement upwards. And this evidence is given under oath and in the genuine belief that such identifications are correct and are unique. That the sequence of points they found are unique and could only have come from that one individual. That no other individual could have that sequence of characteristics in agreement.

My main concern with your posting was not with you seeing a 15 point close non-match but was with the fact you found no discernible dissimilarities.

If you can have two prints in which you have 15 points in close agreement with none in disagreement this potentially undermines hundreds or indeed thousands of convictions that are being determined by experts finding less than 15.

You claim that numbers are irrelevant. But you and your colleague are the only ones who have apparently seen this close non-match, and it is yourself that is claiming it had 15 in close agreement with none in disagreement. You admitted to Mr McKie that you might have been mistaken but in your posting yesterday you are now very sure what you saw.

Furthermore, the question isn't why hasn't this been published? The question is why hasn't anyone else tried this before? (Also Rhetorical)

We have to disagree over this. If the print you saw is as you claim I genuinely believe you have seen one of the single most important prints in the entire science of fingerprinting. But it’s okay, keep it to yourself. You know what you saw. What would be the point of anyone replicating this so-called test for themselves if they were never able to publish the results?

I am thinking we need to leave it there 16 points. It is such a shame that you and your distinguished colleague have felt the need to suppress what you saw.

[Les Bush]

As always the good advise we receive during our time as forensic scientists is never communicate when you are angry or upset as the language and the tone can be interpreted differently to the intention of the message. '16 points' to me is yet another annonymous poster whose explanations doesnt have the basis of either sound reasoning or logic. In any other scientific discipline they would similarly expect that if a credible example existed about how their scientific principles could be eroded they too would seek evidence to investigate the claim. Without the evidence the claim by the anonymous poster remains as colourfully described by the previous posts. Why do we create a fuss when (they) use this site for unsubstantiated postings; because we are experienced examiners, we know our science and the variables upon which our determinations are made. Is this site a democracy to use the right of freedom of speech ect, or is it as the Home page invitation states a site for fingerprint examiners. Also we know our scientific principles and will remain sceptical about the extent and degree of reliability claimed by statistical modelling. While I always support research there must be a proper balance between the reality of our truth and how it can be measured. Regards from oz. Les

[David Johnson]

Les you said

'16 points' to me is yet another anonymous poster whose explanations doesn’t have the basis of either sound reasoning or logic.

Les I think you need to look up sound reasoning and logic because that is exactly what '16 points' statements contain. Some of the posters in your camp seem to be satisfied with saying that there is no way that these fingerprints could exist and therefore we have nothing to worry about.

I believe that I have seen the fingerprints that 16 points is referring to I would have to agree with the statements that he has made about them. The two prints were inked prints with one being the tester and the other being the hit from AFIS. Now while these prints may not have stood up to ridgeology. (especially if you know ahead of time that they are from a test) People will of course scream up and down about how this ridge was to short or that one is off by two microns. In my opinion a latent from the tester would have been very very difficult to tell apart from the known in AFIS and probably would have generated erroneous Identifications in many cases.

And the existence of close non matching prints even out to 16 points does not invalidate the principle fingerprints are a very powerful tool to be used to identify people. What you should be asking yourself is not "does this 15-16 point match exist" but how many more exist and is it cause for concern. Do I need to adjust my threshold when searching very large databases in order to compensate for these close non matching prints?