Last week
we looked at a hypothetical scenario where an examiner "pushes the envelope" and makes an ID that they really were not able to make (they didn't feel completely sure) but the answer happened to be correct.  I received several comments about this not being a type 1 error because the answer was not incorrect.  I had off-line discussions on this subject, and still feel if the hypothesis is "sufficient uniqueness for individualization" and the examiner says "yes" when they should have said "no" then it's a false positive for their ability level.  I'll speak more to that issue in St. Louis.

This week
Gerald Clough brings us the results of discussion and his thoughts on the importance of objectivity during analysis and comparison:

___________________________________

Of Errors Great and Small
An Examination of Rigor in ACE-V

Gerald Clough
Caldwell County (Texas) Sheriff’s Office


The following analysis has grown out of discussion on The Detail chat board, private correspondence with Kasey Wertheim and an ongoing thought experiment in search of rational explanations in a particular class of erroneous identifications. Like any such developing thesis, it is easy to forget that many or most readers haven’t been along for the whole ride and will become wrongly oriented. But I’ll try to present what led to the radical-sounding, but not really so radical, notion that we frequently work a poor process and that particular modern conditions have brought trouble home to roost.

Many of us, I think, share with me a wonder that so many of examiners with above-average credentials have become principles in erroneous identifications. It’s not just that one or another excellent examiner had a bad day. Their conclusions were confirmed by other examiners with more than adequate training and experience and were, in at least one case, confirmed by defense-side examiners who can hardly be accused of bias, unless it be a bias in favor of other examiners. An additional conundrum is why, once the errors are revealed, other examiners appear to readily conclude otherwise.

I have come to believe these errors were not the result of the sort of expectation bias, the significance of which we debate. I do not believe they were doing sloppy, lackadaisical work. They are not that sort of folks, and the cases were sufficiently important to compel close attention, even from a slob. And I certainly don’t believe they were working a conspiracy. (I include in these assessments the examiners of the S.C.R.O. Regardless of their organization’s refusal to admit error, I hardly think we can accuse them of being haphazardly trained and lacking experience. In fact, I believe my thesis, if correct, may explain why they claim the dispute is a matter of difference of opinion.)

Follow along, now, as I present some ideas.

My undergraduate and graduate degrees are in mathematics and computer science, respectively. So, for me, an equation is a statement that is true by definition. It expresses a genuine relation. "2+3=1+4" is a true statement.  "x-1=7" is a true statement when the value of x is 8. Those true statements *define* the values used in both sides. To say that 2+3=1+4 is, to be sure, an inadequate way to define the values represented by the characters "1", "2", "3" and "4", but, with a few conditions, that is its effect.  Given the standard base10 numbers, the statement x-1=7 absolutely defines "x". When we say x-1=7 is true, we presume the value of 8 for x. An equation might be called an expression of an absolute conclusion reached through study of the number system, the representation of values, within which it is true. It’s the end product, but in mathematics and latent print examination, the interesting part is the study of values, rather than the plain statements of conclusion.

An unknown latent impression sufficient for individualization - we may refer to it as “U” - has a value. It shares a value, “u,” with it’s source, an individual and as yet unspecified, portion of human skin. Similarly, an known impression, “K,” has a value, “k,” the value of a portion of skin known to be it’s source. The statement u=k is an equation when the two values are the same, when they have the same source. Of course, we initially know only that u may represent any one of a very great many values. We don’t yet know that value. We only know that somewhere in the world there is or has been a common source.

Our field is grounded in the assumption that for every sufficient portion of friction ridge skin, H, there is a unique value, h, and all sufficiently detailed impressions of that skin portion have the same value. We may examine various impressions of different portions of human skin, those impression having values, h1, h2, h3, . . . If we judge that hn=hm, the value of one impression, hn, being equal to the value of another impression, hm, we conclude that they have a common source, a unique portion of human skin, H. If we judge that hn<>hm, we conclude that they are differently sourced.

When we begin a task of comparison, the relationship between a latent impression (an unknown) and an inked impression (a known) cannot be stated as an equation. But we could properly express our problem as a logical expression. In a sort of computer pseudo-language, we could use the statement “unknown.EQUAL.known” or as a logical function EQUAL(u,k). These statements take on one of two binary values, TRUE or FALSE, depending on whether or not u is equal to k.

(Let me exclude, for our purpose, the case in which the known has no value, being insufficiently detailed to allow identification of the portion of skin with which it might share a value. Attempting to use it would produce the equivalent result, typically an *error* message, of trying to operated the logical function SUM(x,y) with x=3 and y=Apple. We need only consider valid inputs to our EQUAL function. For the mathematically disinclined, a function is a just a machine into which you can feed the sort of things it is designed to process and get back whatever that machine is designed to produce. If you feed it other than what it was designed to accept, it doesn’t run. The human functions we operate in latent print examination produce useful results only when fed the proper arguments, impressions that can be individualized.)

Now we drop down a level to the analysis of the features of the impressions, the details, specifically Level 2 details. As a part of our analysis of an unknown, we operate a human function we will call C(f). The function, C, takes as its argument, its input, a feature observed in the unknown impression. We will call that feature “f”. So, when we observe f and use it as input to our human function C, we return one of two values, TRUE CHARACTERISTIC or NOT A CHARACTERISTIC.

I heard the screeching of mental brakes, just now, but carry on with me. I submit that C is a binary function, that there can be no inconclusive result, that you can either conclude that it is a characteristic, or you cannot conclude that it is a characteristic. (Okay, is that better?) That is the sticking point, that word, “cannot.” Note that the function I called “C” has only one argument, a feature in the unknown impression.

We’re now ready to deal with the central question: Can there be a tentative or conditional conclusion that withholds judgment until the known exemplar is analyzed?

To hold that some features of the unknown cannot be judged to be genuine characteristics without reference to the known is to say that there must be two different functions to evaluate features in the unknown. There must be C(f) to operate on all features that may be judged without reference to the known - and there must be D(f,k) to operate on all features that might or might not be genuine characteristics but that might be decided by using k, a feature of the known impression. That would be to say of a given feature, “Well, it might be a characteristic, and I shall see if it appears in the known and then declare it genuine, if it agrees.”

I would hope that the tone of such a statement would rub an examiner the wrong way, but would be not so irritating as, “I know I said that was artifact, but now that I see the known has a characteristic that corresponds to my ‘artifact,’ I declare the artifact to be genuine.” It rubs me wrong. I do not accept that there are two entirely different functions for evaluating features in the unknown, one that only looks at the unknown and another that considers a feature of the unknown in the light of the analysis of the known.

I believe that the whole foundation of comparison is constructed of absolutes and that we cannot base an absolute conclusion on shifting conclusions of the existence of the very things we are using to make the comparison. I see it as the most blatant of biases to imagine that I cannot say this feature is a characteristic when I analyze the unknown, but that I CAN say it is a characteristic when I reference a known as a potential match. How can I defend such a decision? If I was unsure, I was unsure, and looking at the known changes nothing in the unknown. It’s still the same ambiguous* - and therefore useless - feature it was on first analysis. If it had not appeared in the known, what then?

What then, indeed? Why then should I not decide ambiguity according to what would EXCLUDE**? If I had otherwise sufficient, matching, clear characteristics to declare an identification, I would, according to the proper procedure, then be considering the second condition of individualization, that there was no genuine disparity between the two impressions. I would NOT use the questionable feature, the feature that I was not able to say was a true characteristic, to declare that the two were differently sourced. That second condition cannot be abandoned. A single genuine and unexplainable disparity would not only be enough to preclude identification; it would be enough to declare exclusion of that portion of source skin.

To declare an initially inconclusive feature to be genuine or an artifact, depending on whether it suits us by its appearance in the known, is a violation of that principle. If I wouldn’t use it to exclude, how could I use it to identify? And if I do now declare that feature to be genuine, what of the next ambiguity, and next, and the next…? I risk ending up with a lavishly demonstrated “absolute” based on a string of biased decisions, but I have likely disregarded as artifact other features that would invalidate my conclusion of identification, features that I might have declared genuine, except that it didn’t fit my prediction.

Obviously, when two impressions can be matched using ample and unambiguous detail, revising the initial analysis of previously uncertain features in the unknown to the status of genuine characteristics does indeed appear to reflect reality. But if a particular feature, an ending ridge lying suspiciously close to a blank or blotched area, for instance, might have been, on initial analysis, real or might have been artifact, it is unlikely that a feature of such ambiguity really allows one to know that one is looking at the true limit of the ending ridge, rather than, say, one or two ridge units short of the end. The designation of the feature as a genuine characteristic is as likely to be erroneous as not, being close to but not quite the characteristic and impossible to differentiate from the last visible unit on a continuous ridge. This assumes, of course, the usual circumstance in which there is insufficient Level 3 detail to settle the issue.

In our examination, the unknown is the subject. It is our anchor to reality. It is the crime scene product, the potential evidence. It is what it is, and there is no other latent that it could be. It is not one of a pool of possible latent prints. We’re not deciding if it belongs in the case. It IS in the case. Maybe we have a known that will be of the same source. Maybe we don’t. But in the beginning it is the latent that provides the only absolute information we have. Until we explore the relationship of the latent to a known, we have absolutely no information about the relationship of the latent to other things. We say we make no could-be conclusions, but that is exactly what we do, if we make a tentative or even absolute conclusion about a known and use that conclusion to generate further similarities in the unknown.

We almost all agree that we carry out our examination with absolutely no regard to any information that suggests the source of the known is the source of the unknown. I cannot reconcile that principle with a process that permits the known to influence the interpretation of the unknown. Using features in the known to validate features in the unknown is, exactly, suggesting the known as the source of the unknown. The fact that the suggestion grows out of the comparison process and not from some other investigative source makes no difference.

It is the nature of the beast, with some exceptions, that of the two, the known and unknown, the latent will be the impression that calls for studied decisions as to the value of features. The known is commonly clear. And it is always true that in the world there exist a very great many fingers, the impressions of which are very similar. AFIS is finding some of those similar impressions, and it seems certain that it will produce them in increasing numbers. (An arguable proposition, perhaps, but it seems a reasonable implication of increasing data and more sophisticated algorithms.)

I propose careful consideration of the notion that we often misuse ambiguous features when comparison of more clearly genuine characteristics in both known and unknown suggest a common source. I suggest that, if we are brutally honest in examining our processes, we will give serious consideration to how often such ambiguous features may have been selected as characteristics without any actual harm to the correctly identified human source but with great potential harm to the process itself when the same practice encounters two very similar but differently sourced impressions.

I don’t think the erroneous designation of characteristics is so willful as this may make it appear that I believe. Given the sorts of suspect pools and files of local and regional criminal knowns available to examiners through most of the history of our discipline, I think it reasonable to expect that when a known was found that was so apparently similar to the unknown that a match was declared, even when some features were improperly associated with characteristics in the known, the likelihood that the person identified was associated with the crime scene and was therefore the actual source of the impression was relatively high.

And my unfortunate but realistic expectation is that, even were a truly erroneous identification to be made under those local conditions, it was likely that the accused was indeed a local criminal and, faced with an inability to prove he didn’t do the crime and an attorney who clearly believed in the prints and explained the poor prospects for fighting the fingerprint evidence, pled guilty to an offered reasonable sentence.

If these improper selections of ambiguous features have indeed crept into our routines through lack of detection and a substantial lack of individual harm, it is not to be wondered at. It would be very much akin to any number of technical and scientific errors that were part of perfectly acceptable procedures in other fields until something changed in the environment or the knowledge base. It is not something that would have been subject to any imagined preventative, such as an arbitrary high minimum point count requirement. The errors are made in the designation of points themselves. It would not be entirely unreasonable to suspect that a high point count requirement could aggravate the problem.

If embedded in common practice, this kind of error seems particularly difficult to detect, since it is less likely to be a beginner’s error. The more experienced and confident an examiner, the more subtle the examination and perhaps the more likely for the examiner to feel able to “make the tough ones.” The trainee is less likely to push beyond clear characteristics into making fine judgments on ambiguous features.

I suspect, too, that the move to holistic examination may have created an environment where such errors are no less likely. The holistic approach is more sophisticated and brings an array of factors into the comparison, but the multidimensional nature also encourages a more creative mindset.

Neither is ACE-V a prophylactic, unless its implied strict sequence is rigorously followed. Such acronyms are useful aids to encourage consistently ordered processes, but they are prone to becoming mantras that substitute for thoughtful practice. I believe the fault will be found in a lack of rigor in the initial analysis of the unknown and in isolating the analysis of the unknown from outside influences. If the analysis of the unknown is left partly open, pending comparison to the known, the order of operations is corrupted, and it is no longer ACE-V - it is ACAE-V. One cannot be even accused of modifying one’s analysis, because the analysis was not declared complete until the result was modified by comparison.

What if this is true, and we fix it? It might be that we would then forgo absolute conclusions of identification of some number of unknowns. I don’t think it would be a large number. I think, though, that we would be citing fewer characteristics in given comparisons, but those we did cite would be easily defendable. I am specifically concerned with coincidentally correct, known-influenced decisions in cases where individualization could have readily been established without resort to questionable features, because those undetected and individually harmless instances become habitual and are exactly the kind of errors that can lead to erroneous identifications when questionable features are used to form the conclusion.

In the face of very public mistakes, I had to consider why I believe I would avoid erroneous identifications when examiners with distinctly better credentials than I have made them. Concluding that they simply “messed up” or had a bad day (the whole unit had a bad day?!!) is the intellectually lazy way out and an implied slander on some very capable people. Part of my suspicion that I might be looking at something significant was my unease with the lack of formal and detailed written analysis as part of the usual examination.

I earlier presented a view of the examination as a human analog to a logical function. But, unlike a logical function as provided in a programming language, the human functions operated in latent print examination are not precisely defined. Each examiner operates a different mind, with different experiences, and therefore operates a set of functions different from every other examiner, all of them intended to produce the same results. It seems perfectly reasonable and necessary that an examination be reported in such a way that the process worked by the unique combination of individual examiner and particular evidentiary materials can be presented and followed and evaluated.

My answer to the above question about why I believe I am not making the same kind of errors as my professional peers and superiors is that I am documenting my analysis of the unknown and my independent analysis of the known, the logical comparison of the two and the basis for my conclusion and that I imposed upon myself the obligation to state just why I believe each individual decision was made properly along the path to the conclusion.

I must admit that I view with some sense of relief the notion that these most troubling erroneous identifications grow out of improper treatment of ambiguous features, because I see it as readily addressable and really a “tightening up” of the fundamental principles of latent examination, rather than a revision of principles. For that reason, I am eager to hear critical comment. For, if I’m off the mark, the problematic errors remain for me undiagnosed and even more troubling. I still reject the easy out of pinning the rap on something limited to the involved examiners.

I must say, in the way of disclaimer, that I do not reject any and all reexamination of the unknown following analysis of the known. It would be foolish and rigidly dogmatic to refuse to deal with a later-realized mistake in analysis of the unknown. Things are not so clean in examination. But I believe it is possible to recognize when such things violate the principles.



* “Ambiguous” requires some definition, in order that we not misunderstand the sort of indeterminate features central to this discussion. A characteristic that could be, for instance, the curving end of an ending ridge lying between two continuous ridges or could be a bifurcation of one of the ridges where the point at which the ridge bifurcates was not captured in the impression is not ambiguous, as the term is used here. In that instance, there is a genuine characteristic, even though the precise structure cannot be known without microscopic examination of the source skin. Ambiguous, here, means that some defect in clarity makes it impossible to say if the observed feature reflects an actual anomaly in ridge structure.

** I use “exclude” as a short form of a statement that excludes as the source of the unknown impression the specific portion of friction ridge skin from with the known impression was made.

_______________________________________________________

To discuss this Detail, the message board is always open: (http://www.clpex.com/phpBB/viewforum.php?f=2)

More formal latent print discussions are available at onin.com: (http://www.onin.com)

_______________________________________________________
FUNNY FINGERPRINT FIND

The FFF folder is empty again!  Send 'em in as you find 'em.

_______________________________________________________
MANAGEMENT CIRCLE

Getting the most out of a seminar

Your organization is footing the bill for you to attend an industry conference or seminar.  How can you make the most of the experience?  The best way is to eliminate distractions and hone your listening skills.  Here's how:

1) Sit near the front,  so you have a clear view of the speaker.  It's easier to be attentive if you're sitting at the front of the room.  If you sit in the back, you may be distracted by others in front of you.

2) Come prepared.  Ask yourself "What will I gain from this presentation?"  Review what you already know on the subject, so you'll be able to relate the information to your own interests.

3) Take notes, but don't try to write everything down.  Don't be consumed by your notes, or you'll be distracted from the overall message.

4) Have high expectations.  You hear what you expect to hear.  If you think the event will be boring, it will be.

-Adapted from "1000 Things You Never Learned in Business School,", by William N. Yeomana via Communication Briefings, February 2004, 800.722.9221, briefings.com.
_______________________________________________________

UPDATES ON CLPEX.com

Updated the Smiley Files with two new smileys

Updated the CloseCalls page with a submission from Charlie Parker in Texas.  Send in your CloseCall today!

_______________________________________________________

Feel free to pass The Detail along to other examiners.  This is a free newsletter FOR latent print examiners, BY latent print examiners. There are no copyrights on The Detail, and the website is open for all to visit.

If you have not yet signed up to receive the Weekly Detail in YOUR e-mail inbox, go ahead and join the list now so you don't miss out!  (To join this free e-mail newsletter, send a blank e-mail to: theweeklydetail-subscribe@topica.email-publisher.com)  Members may unsubscribe at any time.  If you have difficulties with the sign-up process or have been inadvertently removed from the list, e-mail me personally at kaseywertheim@aol.com and I will try to work things out.

Until next Monday morning, don't work too hard or too little.

Have a GREAT week!