Complexity and the Comparison

[Boyd Baumgartner]

Complexity, Level of Association and Strength of Fingerprint Conclusions - TRIPLETT - 2016 PROOF.pdf

Michele has a new paper coming out regarding the complexity of comparisons and how that relates to the strength of a conclusion and was wondering if anyone has any thoughts on the topic (pre or post reading)or attempted to measure such a thing outside of a likelihood ratio.

The TL;DR version is that a more complex comparison could lead to a diminished level of association for an ID and that this could be communicated in a verbal scale. A real world example could include agencies that use inconclusive as trending ID but lacking sufficiency, the dreaded AFIS inconclusive, or just moving towards a questioned document style scale of association.

Do you think this more clearly and transparently articulates our findings or do you think it waters down conclusions to the point of being a detriment to the trier of fact? Will we end up with 10 versions of what 'complex' means depending on who is on OSAC or can we apply quantitative techniques to get a more standardized measurement?

Here's a paper from 2014 that seems to allude to the second idea.

Discuss....

[LPE123]

Unlike DNA, we who work in fingerprints do not have a statistical model on which to calculate likelihood ratios, uncertainty, or error rate.

By the way, "uncertainty," as I understand it, is a term used to communicate a specific scientific calculation. "Error Rate" would be used to refer to the examiner's opinion outside the scientific realm of "uncertainty." Would anybody like to comment and help me understand the difference?

That aside, I know that in my own mind, I still cling to the old dogma that I was taught a decade or so ago. A comparison either results in an identification, an exclusion, or a result of inconclusive.

I know that Rule 702 allows an "expert" to testify to an "opinion." Therefore, I feel resistance to the idea of "strength of conclusion," since we lack any validated, calculable model on which to quantify or qualify that "strength."

It is either him, it was somebody else, or I can't tell.

[Boyd Baumgartner]

I don't think you're alone in your thinking of Is/Isn't/Can't Tell with regards to prints, but even this line of thinking is relatively new to the discipline. It used to be strictly positive match, or negative. Despite it recently being rescinded, things were this way in part, thanks to the IAI's Resolution 1979-7 and 1980- 5 which state:

Resolution 1979-7:
“THEREFORE BE IT RESOLVED that any member, officer or certified latent
print examiner who provides oral or written reports, or gives testimony of
possible, probable or likely friction ridge identification shall be deemed to be
engaged in conduct unbecoming such member, officer or certified latent print
examiner as described in Article XVII, Section 5, of the constitution of the
International Association for Identification, and charges may be brought under
such conditions set forth in Article XVI, Section 5, of the constitution. If such
member be a certified latent print examiner, his conduct and status shall be
reconsidered by the Latent Print Certification Board….”

Resolution 1980-5: (Amending Resolution 1979-7)
“THEREFORE BE IT RESOLVED that any member, officer or certified latent
print examiner who initiates or volunteers oral or written reports, or testimony
of possible, probable or likely friction ridge identification, or who, when
required in a judicial proceeding to provide such reports or testimony, does
not qualify it with a statement that the print in question could be that of
someone else, shall be deemed to be engaged in conduct unbecoming such
member,…”

Throw in a little infallibility bravado, and I'd argue that this 'no third way' thinking lead in part culturally, if nothing else to the Mayfield error. Going back as far as 1995's Ne'urim symposium Christophe Champod and Pierre Margot argued:

"In fact, there is no logical reason to suppress grey levels between white (exclusion) and black (identification). Each piece of evidence is relevant if it tends to make the matter at issue more or less probable than otherwise. Dactyloscopic evidence can be combined with other (forensic or not) and fitted into the particular context of each case."

"The refusal of qualified opinions in dactyloscopy is a policy decision [even if the distinction of the arguments (policy or scientific argumentation) is not so clear in the literature]. We think that the absence of extensive statistical data on fingerprint individuality is the main reason to prevent entering in the grey area."

Even Subsection 8 of Resolution 2010-18 (First Link) states that a statistical model is not to be used in absence of the opinion of an Examiner. So, rather than the 'all or nothing' of waiting for a validated probability model which isn't going to do any heavy lifting anyway, does it make sense to move towards a standards based qualified conclusion scale? There's already a trend towards qualifying exclusions through the use of standards and a trend towards distinguishing 'I can't find it due to the ambiguity of the orientation, anatomical source and focal points' inconclusive vs the 'this would be a match if there were just a little more data' inconclusive. And if people are already moving towards this line of thinking, why not create a scale that links these trends?

[Simon Bunter]

A very eloquent paper with some well articulated arguments. I think when likelihood or probability is mentioned to many Fingerprint Experts they automatically think it entails expressing an identification in a numerical format such as, for example, 'a 1 in 51,583,902 chance that the mark was made by someone else' (similar to DNA reporting). This doesn't have to be the case as many other forensic sciences use a scale of support without a numerical basis, e.g. CCTV imagery analysis etc. As Ms Triplett explains in her hospital example, these types of scales are not simply at the discretion of the examiner but based on agreed criteria for each category of the wording scale.

In the UK there is a move towards a three result reporting system for sufficient marks; exclusion/inconclusive/identification. With exclusion and identification basically 'conclusive' results at the extreme ends of the scale, the 'inconclusive' result seems to cover everything in between, which makes it too wide/vague to be meaningful. In essence, 'inconclusive' appears to cover every strength of opinion between "I'm 99% certain that it was NOT him" to "I'm 99% certain that it WAS him". Due to this, I think sometimes Fingerprint Officers might feel forced to give a result of 'identification' even if there is an aspect of the comparison that causes them some cause for concern. A worded scale would allow a Fingerprint Officer to provide a slightly less than conclusive result if they felt it was warranted, e.g. 'very strong support' or similar.

In my experience, the current 'conclusive' method of reporting can sometimes be disingenuous or greatly exaggerate the strength of the match. I have examined fingerprint evidence in many cases where, in my opinion, the strength of the 'conclusive' identification had been exaggerated. For example:

-'Identifications' reported as containing '16 matching ridge characteristics' where considerably fewer were demonstrable.

-'Identifications' reported as containing 'no unexplainable discrepancies' where several differences were noted but no cogent reason could be given by anyone, only a list of 'possibilities'.

Where there has been disagreement in such previous cases, in lieu of an agreed fingerprint scale, I have explained my findings using a mixture of layman's terms and, where appropriate, reference to a generic forensic scale of support, e.g. 'strong support' from a scale of "neutral, weak, moderately strong, strong, very strong, extremely strong". As there is no agreed fingerprint support scale, however, it is sometimes difficult to select the 'correct' level of support and then justify the reasons for that selection.

If the profession is to move forward, embrace modern forensic thinking and make the most of the countless research publications in relation to improving the reliability and transparency of fingerprint evidence, the type of scale suggested by Ms Triplett would be a good first step.

[Les Bush]

Hello to all from down under, all the best for 2016, even though I have retired it is still in my forensic DNA to contribute to discussions and ideas relating to Fingerprint Science. A fingerprint is a two dimensional pattern made from a three dimensional source and one that is a unique biologically produced arrangement of features. An expert should be able to explain how a pattern of those details has been attributed to one person. The questioned evidence of a two dimensional latent is a geometric arrangement/pattern having shapes, spatial data and degrees of clarity. An expert should be able to explain the quantity of data visible and observable in the latent print. The process of comparing any latent to an exemplar from a known source skin should follow a logical and scientific method. An expert should be able to explain the repeatability of the examination process including visual capture of details and separation of the development hypothesis and evaluation hypothesis leading to an examination conclusion. Armed with those quantifiable details, a defined examination process including visual aids to demonstrate repeatability, and the knowledge of the fundamental principles of biological uniqueness, the expert should be able to present intelligent and rational testimony to a court. However, the rub is this, because of international and even national diversity of agencies and generational differences between dogma trained and academically trained experts the likelihood of achieving consensus about a universal approach toward processes such as comparisons and conclusions is very unlikely. That is not being sceptical, just pragmatic and realistic. The starting point for change has to be the new initiate into Fingerprint Science with the learning and experience of that person shaping both them and the Science into the future. As an example of this would be the post generational changes to the fingerprint system in the UK following the McKie case. Is it still a divided house or one that is pushing forward? And as an example of Complexity using the scale system of this thread, was the McKie print a complex latent, or an easy exclusion? An expert should be able to explain the difference while making a clear distinction. Cheers from oz les

[Boyd Baumgartner]

I don't know how I missed this earlier but it's David Kaye's blog and it discusses this very topic through a critical look at the Army Crime Lab's move away from the Individualization/Identification terminology. There are some other good entries like a look at the Miami Dade Study too. Check it out.

Marching Toward Improved Latent Fingerprint Testimony at the Army's Defense Forensic Science Center