CLPEX Chat Board

Les Bush wrote:Once again perhaps Gerald can help us here to clarify which of the fundamental principles requires validation. Regards. les

For a long time (and still) after serious criticism arose, the language used has been not very precise. There was talk of "validating" "methodology." Sorry. I can't tell you what that means. I know, though, what criticisms were (or at least now are) intended. So that we don't talk at cross-purposes, I'll state (again) how I view this. You can't talk about "validating" the human process of observing and interpreting. That activity does not require any theory, and it reaches no conclusions, so there's simply nothing to find valid. You can talk about one or another person interpreting more consistently correctly. You can even characterize how well a group of people do it on average, but that tells you nothing about an individual examination, except to tell you that you ought to recheck, because humans can make observational and interpretive error. So, if we're to talk about the same thing, you have top understand that validation has NOTHING to do with how well an examiner performs. Nothing. This is very important, because it is a great error to in any way attempt to use the results of latent print examinations, the historical record, to achieve scientific validation.

When I speak of "validation," that's what I'm talking about, scientific validation of this principle. Note that it makes no mention at all of HOW the principle is applied.

Human friction ridge skin has observable physical characteristics that occur with sufficient variability that the likelihood of two different portions of such skin cannot be differentiated is specific probability.

It can be stated in other ways, but the key points are:

It's skin, not impressions of skin. Impressions are one means of observing skin, but we are talking about skin. Validation is done with ideal skin data. As a practical matter, one might indeed use scans or impressions to code data for validation studies, but it's not necessary to do so,

The principle cannot be the any portion of such skin is unique in its observable characteristics. Such a theory is a frank logical failure. (If someone's been following the discussion for weeks and doesn't get this one yet, don't stay with it. Read the horoscope column or something. Avoid science. You're not up to it.)

Since we are not claiming absolute uniqueness, it's possible to be unable to differentiate two portions of skin by observing the characteristics. Yet, we need to make absolute statements as our conclusions from doing this observation and interpretation during an examination. If we can't declare absolutely that one and only one person could have a portion of skin that we could not discern as different from the portion observed, the only precise statement we can make is to state a specific likelihood of that event.

If we wanted to go forward, fat and happy, making absolute conclusions, we would have to state the principle very differently. "The is some quantity of skin that is absolutely unique in its observable features." That CANNOT be proved. Not HAS NOT been proven - CANNOT be proven. That principle fails, and conclusions of that sort cannot be validated. Further, they are frankly untrue. No experiment needed. Logic alone does the trick. As a principle to be applied to produce scientific conclusions, that's a complete dead end. It is NOT a matter of "not-validated" not implying "invalid." It is true that not being validated does not imply invalid. But the theory still has to be logically consistent with reality. Absolute uniqueness is NOT. If you with to claim that any portion of friction ridge skin is unique in its features, you MUST hold that the number of variations if INFINITE. You cannot produce an infinity of variations without an infinite set from which to draw arrangements. And, so far as skin is concerned, that cannot exist in nature. Even if we weren't talking about just visually observable features, if we were talking about every possible aspect of skin - cells, DNA, etc., the variations would not be infinite, and therefore, no example can be claimed to be absolutely unique. So it's not just a matter of skin being unique but humans not being able to observe the uniqueness. You can't say "more unique" or "less unique." That's just a weaseling way of avoiding probabilities.

That's all there is to validation. Nothing about biases or optics or judgment. It's just the validation of the theory of physical reality that no one doubts will be validated, but it never has been validated. That is seems to work doesn't count scientifically. And if you want to claim scientific conclusions, you simply have to work from a validated theory. Otherwise, when tested as science, you're just guessing. No matter how often you're right. No matter how sophisticated the guess. It's a guess. No matter how high you set your threshold, it's a guess. And when it's an absolute statement, it can never be scientifically or logically valid.

It would be wonderful indeed if the principle that some portion of skin is absolutely unique in its observable features could be validated. But that's simply not possible. It's simply not true. Very, very, very small likelihood does not make it true. Clinical success does make it scientifically true. That's what this is about, acceptable scientific validation. You can proceed without it, but your conclusions will ultimately be rejected as based upon not validated science.

Validation is necessary, if one is to claim scientifically valid conclusions. In terms of practical examination, it is the most trivial part. Or it would be, except that it inevitably expresses in terms of probability, simply because we are dealing with a finite number variations in skin features. The issues of how humans observe skin characteristics is of arguably greater importance, because that's where we get our data. It's where error can be made. None of that is subject to validation. ACE-V isn't a matter for validation. You cannot state a proper theory that a human being under specific conditions will correctly interpret impressed features. Again, you might come up with error rates, but they have nothing to do with any individual examination, other than maybe a general sense of how suspicious you might be about interpretive error.

You know, the principle IS valid. No one doubts that. And it's going to be validated. Depend on it. No avoiding the issue is going to stop that. You're going to end up with a scientifically validated discipline. Congratulations. And when that principle is proven, the precise meaning of a correctly interpreted impression will be known.

Hello Gerald,

Thanks for your thoughts and I agree there does seem to be a lot of repetition creeping into our discussion. Perhaps we are not meant to find agreement sufficient to begin the process of designing a validation test. It does appear you have strong views as highlighted by the upper case text. I can also see that Kasey has plonked in a new posting and it relates to the purpose of this thread as being 'recent news' so I'm assuming he wants us to finish up or start a separate thread. Before I leave you made the call that there is a finite limit to the variations in fingerprints because infinite cannot exist in nature. In that sense your view is there are bounds to the diversity in biology and that random coupling of DNA will still only have a finite set of reproductions. Okay, lets look at a study of monozygotic twins their fingerprints are unique. Lets also look at the millions of ten print collections globally and the absolute absence of proof of duplication with the consequent evidence of uniqueness. Even Glenn in his last posting agreed that uniqueness in fingerprints can be achieved by ten prints. What it is we practice in our analysis of fingerprints establishes a truth value that one person only is reponsible for reproductions of their fingerprints. We do it for both tenprints and latents since the same principle applies. This remains a practical reality as no evidence or proof exists that our principle is incorrect or invalid. I agree we can make some comparisons with other sciences in how they analyse data to form a conclusion but we should not have to compromise the range of variables that make fingerprint science a distinct discipline. I think the answer as to how we meet the expectation of scientific validity will come from within fingerprint science practice given the expertise to determine the truth about fingerprint patterns is owned by those who study them. Thanks also to Michelle and Steve for their recent input. Cheers from oz. Les

I don't mean to sound simple but isn't method validation just - error assessment?

What I mean by error assessment is what percentage of error occurs when applying a particular method?

What makes this anymore scientific than the real world (rather than inside the lab tests) examples of errors?

raybie wrote:I don't mean to sound simple but isn't method validation just - error assessment?

What I mean by error assessment is what percentage of error occurs when applying a particular method?

What makes this anymore scientific than the real world (rather than inside the lab tests) examples of errors?

Well, Les is right. We're getting repetitious. But briefly, validation, as I've been speaking of it, has nothing to do with errors in observation or in how a less than perfectly impressed detail is interpreted. Validation has to do with a scientifically rigorous proof if the fundamental theory that the observable features of friction ridge skin can be used to powerfully discriminate among individuals, which really means how frequently a specific quantity of detail in agreement will be found in the population. Validation, then, is study of reality, to the end of validating its application to the practical tasks of latent print examination.

You cannot (taking validation as I use the term here) validate a method. A method is not a theory or reality. You can, however, two two things with method. One is study overall error rates among some defined population of examiners. I seriously doubt the value of that, because it says nothing about any particular case, except that it maybe alerts someone to the possibility of human error. And we already know that's a risk. The other thing you can do with method is study specific kinds and sources of error. The only kinds of error worth talking about are errors in observation (just plain being mistaken about what was there to see) and error in interpretation (mischaracterizing something when it required some expert judgment to say just what actual skin feature is represented by the observed impressed detail). These are valuable studies, because they tell us how we can better prevent those errors. They tell us what sorts of bias are at work and which matter and which don't. And they tell us how better to structure review of casework. They unfortunately do not tell us how to achieve scientific certainty that none of those errors were made and led to incorrect conclusion.

That's a revealing difference between error study and validation. When you have done validation, you have satisfied to an acceptable scientific standard why what you're doing gives you results that apply to reality, and it allows you to speak, again with scientific certainty, that, if you observed and interpreted correctly, your conclusive statement is true.

But since we are indeed getting repetitous and have drifted far outside the "Recent News" topic. I suggest you go back and read the earlier material from this and other threads on the subject and maybe my two-part essay in The Detail. And I will no longer continue this subject in this topic.

And, Les. I want to think a bit about what to say about "uniqueness" and will pitch it in a new thread if I have time. I am seeing that this "unique" thing is kind of a key to some of there concepts.