Show me the print - a group exercise

[Boyd Baumgartner]

Edit: Poll added (you can only add a poll to the first post in a thread)

I believe Pat's old nickname is 'show me the print'. With all the talk of ground truth vs casework, GYRO, consensus, standards and whatnot I figured I'd channel that 'show me the print' spirit and try and do something different by taking us out of the philosophical and in to the practical.

Here is the scenario: You are asked to consult on an old film case. You are given an 8 X 10 (300 ppi) photograph of a latent and asked to analyze it. Utilizing the current tools and techniques you will apply GYRO and analyze the print linearly (plot minutiae and confidence via GYRO before being exposed to the known print)

All submissions will be anonymous and linked to a unique ID that you create.

I will post part 2 next Monday.

If you are using an older/mobile browser, here is a direct link to the form with the instructions
https://goo.gl/forms/3J3muzyQQfMbzWsP2

Here is the image at 300 ppi at 4X5 in jpg format. The 8X10 lossless .tif can be downloaded through the link in the instructions

Latent-300-ppi-4x5.jpg

[Boyd Baumgartner]

da890830-ea2d-4d7c-9be9-1fcb38a8b14f
b5253fca-e5a6-41ae-9c24-9bdc20b8419e
64b85233-360a-4fe2-b6ea-0a464751964f (edit: I received this. thank you)

Could you please try re-uploading your .psd files. I updated the script that allows google drive to be used as a dropbox to work more efficiently and give you a confirmation.

https://script.google.com/macros/s/AKfy ... F49Wg/exec

It'll look like this:

dropbox.jpg

[Pat A. Wertheim]

I believe Pat's old nickname is 'show me the print'.

I regret delaying my reply to the above quote, but I have to set the record straight. It was at the 1984 conference of the Texas Division of the IAI where I first heard the term "Show Me The Print!" in a talk given by Axel Hansen, a true old time latent print examiner. Axel joined the Texas Division in 1949 (I was still in diapers) with Member # 264. Axel was what we in Texas would call "a character." He was loud, outspoken, gregarious, exaggerated in every mannerism, and he would go to extreme lengths to help another member with a problem. I learned from Axel not to speculate on what I might do with some hypothetical latent print without first seeing the actual print. His words became a part of my lectures as I have tried to pass on the wisdom of my mentors. Axel died over 30 years ago, but he lives on in my memory. For the fact that "Show me the print" has become a part of our vernacular, credit goes to Axel Hansen.

Axel lives!

[Boyd Baumgartner]

Part II includes comparing the impressions and arriving at a conclusion. Continuing with our theme of using the latest methods and techniques, we will be utilizing the draft version of Standards for Friction Ridge Examination Conclusions published by OSAC a couple months ago.

I will post the results of part II in one week and we can begin the discussion.

Here is a half sized jpg showing the latent and the known side by side. The link to download the lossless .tif file and complete instructions are included in the form.

Comparison.jpg

If you have trouble viewing the embedded form, here is a direct link:
https://goo.gl/forms/xe3iWH4Byi0P4bnp2

[Boyd Baumgartner]

First, the results of the Analysis. There were a total of 8 participants.
I've provided:

• an animated gif that cycles through all the markings
• a composite markup
• minutiae with at least three markings in the most dominant color. Equal numbers of colors are represented by split colors
• the answers from the form

Latent-GYRO-GIF.gif

Latent-GYRO-Composite.jpg

Latent-GYRO-Consensus.jpg

[Boyd Baumgartner]

Now for the Comparison. We only had four participants and only three submissions of markups (one person only filled out the form).

I present:

• An Animated GIF of the three markups
• A composite markup (the composite layers are reduced opacity so that overlapping minutiae appear stronger)
• The responses from the form

Comparison-GIF.gif

Comparison-Composite.jpg

Comparison-Questions.jpg

[Boyd Baumgartner]

I've added a poll to this thread so that you can provide your input on a conclusion based on what you've seen. You'll have to scroll up to the top of this thread, as that's the only place a poll can be placed.

You don't have to be a registered user to vote.

[Boyd Baumgartner]

I was asked to put the before and after markings on the latent of the people who submitted. This person was interested how the original GYRO markings might have moved if at all.

Here they are. The Comparison markup comes second in the animation and is reduced to 60% opacity. The file names are the first part of the Unique ID of the person.

f6a1e6a4-AC.gif

b5253fca-AC.gif

6587696f-AC.gif

[Boyd Baumgartner]

Ok, so let's get the ball rolling on a discussion.

To me, the interesting questions and their tangents are something like:

Is this a complex comparison? What makes it so/not?
If this is a complex comparison AND it's an ID, is it as strong of an ID as say a 10P to 10p ID? Why?
Should the strength of an ID be reported as well for transparency's sake?
What role does objectivity play in the conclusion?
Do minutiae introduced after the exposure to the known diminish their value? (the GYRO paper seems to think so)
If you need the orange minutiae to tip you over the threshold for sufficiency for an ID are you biased?
If you consume all the data in order to ID the print, does it diminish the weight of the conclusion (ID more correctly weighted as 'Support for ID')
Should objectivity be a standard for a conclusion?
If you have to alter your original GYR minutiae from the analysis, should that really cause them to be orange?
Is GYRO a sufficient tool given that it doesn't mark sequence, distortion, or differences in data between the prints?
How does one reconcile GYRO's notion of confidence in minutiae with subjectivity of minutiae?
Given that the comments in the Analysis were remarkably consistent regarding the presence of overlapping ridges, can an ID be aggregated from two fragments of what is arguably the same print?
if people render a conclusion outside of the majority, should that be considered an error? Should they be held accountable?
Does your agency utilize a process similar to this where you render a consensus conclusion? (AFIS Inconclusives, Complex Prints, etc)

Feel free to tackle any of these or add your own.

[Pat A. Wertheim]

Is this a complex comparison? What makes it so/not?
Yes, this is a complex latent print. The factors contributing to this analysis include the probable double-tap, low contrast in a large part of the print, and the need to approximate the positions of many of the ridge endings/bifurcations.

If this is a complex comparison AND it's an ID, is it as strong of an ID as say a 10P to 10p ID? Why?
I was not comfortable calling this an identification. I reject the philosophy of more likely than not, or less likely than not, because I have seen far too many close non-idents in forty years of latent print comparison experience. Especially in the AFIS realm with data bases of hundreds of millions of fingerprints, the risk of error is to great to gamble on a “probable” identification.

Should the strength of an ID be reported as well for transparency's sake?
I believe the risk of misunderstanding or misinterpretation of a “probable” conclusion on the part of investigators and jurors is too great. It is better to just say “inconclusive.”

What role does objectivity play in the conclusion?
I believe the role played by biases is overstated in most cases, but with an unclear latent such as this, I could accept that contextual and confirmation bias might influence an examiner. That said, I have seen some examiners become more cautious in comparisons and verifications in order to avoid the effects of bias, while I have seen other examiners barrel headlong into a comparison oblivious of bias.

Do minutiae introduced after the exposure to the known diminish their value? (the GYRO paper seems to think so)
While I understand the idea that the inked can lead the latent during a comparison, I believe that an examiner can compensate by using due caution.

If you need the orange minutiae to tip you over the threshold for sufficiency for an ID are you biased?
I am not opposed to using the orange minutiae, but the weight of a point varies. The green points have the most weight, the yellow slightly less, the red still less, and the orange the least of all. It is not like a balance scale where sixteen little brass weights, each weighing exactly one ounce, make a pound.

If you consume all the data in order to ID the print, does it diminish the weight of the conclusion (ID more correctly weighted as 'Support for ID')
I frequently hear another examiner say, “If I just had one more point . . . “ and I have often felt that way myself. If I consume all of the data? I don’t know – show me the print.

Should objectivity be a standard for a conclusion?
The process, correctly followed, is objective, but the decision is subjective. “Subjective” does not mean dangerous or wrong. Somewhere along the way, the word “subjective” has come to take on a negative connotation. Nothing could be further from the truth. We make millions of correct “subjective” decisions every day. If an examiner is trained, experienced, and tested for competency, that examiner's subjective decisions should be accepted as accurate and reliable, unless proven otherwise.

If you have to alter your original GYR minutiae from the analysis, should that really cause them to be orange?
Each point either adds or subtracts confidence from the conclusion. It is the total weight of the sum in the final evaluation that is important.

Is GYRO a sufficient tool given that it doesn't mark sequence, distortion, or differences in data between the prints?
GYRO is a tool sufficient for most comparisons. But any tool box should have special tools for special cases.

How does one reconcile GYRO's notion of confidence in minutiae with subjectivity of minutiae?
I see no conflict between using subjective analysis to assign value representative of the confidence level for each point.

Given that the comments in the Analysis were remarkably consistent regarding the presence of overlapping ridges, can an ID be aggregated from two fragments of what is arguably the same print?
In deciding whether aggregation is appropriate, I would look at Ashbaugh’s analysis protocol. Does the matrix appear to be the same? Did the development medium react to the matrix the same? Are the biological factors of the ridges (ridge width, texture of level 3, etc.) the same?

if people render a conclusion outside of the majority, should that be considered an error? Should they be held accountable?
Only if they can be shown to be wrong. If the difference is Inconclusive versus either ident or exclusion, than saying it is an error maybe too strong an indictment.

Does your agency utilize a process similar to this where you render a consensus conclusion? (AFIS Inconclusives, Complex Prints, etc)
No.

[Boyd Baumgartner]

Should the strength of an ID be reported as well for transparency's sake?
I believe the risk of misunderstanding or misinterpretation of a “probable” conclusion on the part of investigators and jurors is too great. It is better to just say “inconclusive.”

Help me understand then, the use of inconclusive over say, no value. It would seem to me that the same scenario exists in the use of inconclusive over no value. Specifically, that use of the term inconclusive smuggles in an implicit interpretation of 'I can't say he didn't do it' by a juror which is, in effect a statement of probability greater than zero, or at least different than 'there was no fingerprint evidence'.

If diminishing an improper interpretation is a goal, then shouldn't it be extended to the greatest possible extent?

[Pat A. Wertheim]

I recall a rape case in which a woman alone in bed late at night became aware of a presence in her bedroom. She switched on the lamp by her bed, but immediately had a pillow thrust over her face as the attacker reached up and turned off the light. The victim and her family were all loops, but one whorl pattern with only 3 points and no clear level 3 was developed on the lamp. In a fishing expedition of possible suspects, I found one person with a whorl pattern that had three points in correspondence within tolerance to the suspect. I reported that to the investigator as an "investigative lead," but would never have testified to any degree of "probability."

In another case I had, a knife used in an assault yielded a fragment of open field with one bifurcation above the handle at the base of the blade. The location was approximately where the right thumb would rest if the knife were held in a forward thrusting manner. Sure enough, there was an open field with a single bifurcation on the suspect's right thumb in a place consistent with the scenario. In addition, there was sufficient level 3 detail to comment on some degree of correspondence. What "probability" could be assigned to that print?

If we eventually have a probabilistic model that can be validated and pass Daubert muster, then I might change my position and be willing to testify to a probability or likelihood ratio as supported by that model. But in the absence of a validated, generally accepted model, my personal opinion is that to communicate some vague degree of probability would be a disservice to the criminal justice system, the court, the jury, and any innocent person who might be unfortunate enough to be tried for an offense. So I repeat,

It is better to just say “inconclusive.”

[anwilson]

Is this a complex comparison? What makes it so/not?
I agree with Pat that this is a complex comparison for the factors he listed. Almost the entire print requires some level of interpretation and reliance on the known exemplar to "place" features that appear consistent.

If this is a complex comparison AND it's an ID, is it as strong of an ID as say a 10P to 10P ID? Why?
Like Pat, I don't think an identification would be an appropriate conclusion in this situation. As I work for an agency that has spent several years working on a complexity scale tailored towards how we do comparisons I have no heartburn saying 1 ID may not have the same strength or risk of error as another. Risk of error just determines how much QA a print may warrant. One that is more complex simply means you may want to get a consultation prior to reporting the result and there absolutely should be a verifier scrutinizing the comparison work performed.

Should the strength of an ID be reported as well for transparency's sake?
I'm going to answer this probably a little differently than intended. I think a valid criticism in our discipline is how we articulate to jurors how we reached our conclusion as opposed to the overall validity of the discipline. I've heard several examiners throughout the years say they shouldn't have to explain to a juror what they relied on to reach their conclusion and that their report should stand on their own. I think that should always be a red flag in the courts. An examiner needs to be able to demonstrate to jurors how they performed their work. Whether that's through a more detailed report, a charted enlargement, etc. can be debated as long as the end result is better articulation of the comparison process what how that relates to the final conclusion.

What role does objectivity place in the conclusion?
Again I'm going to go a slightly different route and piggyback off something Pat brought up. Subjective has taken on a weird negative connotation in our discipline. This is true for objective as well. There seems to be a misunderstanding going around that if a computer gives you an answer that result is somehow more objective than doing a visual comparison (seen by some as solely subjective). Every single area of science involves subjectivity. Subjectivity simply means a human is interpreting information available to them and reaching a decision based on their observations while objective means something occurs in nature and is free of a person's belief/interpretation. There's this big push to utilize statistical models in our field because some feel this will make the discipline more "objective". An important thing to keep in mind is that statistical analysis is not a thing that occurs in nature. Every single model requires a human to interpret what variables are important for their models and what variables don't need to be addressed. By it's nature, statistics is subjective which is why it's an applied science and not an exact science. Further, humans have to interpret what the results of the model mean which requires their interpretation. In terms of objectivity in our conclusions, that goes to the clarity of the information. If a ridge ending is simply a ridge ending in a print with no distortion that's not my interpretation that is exists, it's just there. My interpretation of what that ridge ending means to a particular comparison and final conclusion is subjective as I am giving a certain weight to that event (and other events in the print) which is outside of nature. Distortion is the biggest driver of how much subjectivity ultimately is involved in the comparison process. More distortion = more interpretation plain and simple.

Do minutiae introduced after the exposure to the known diminish their value?
For prints with no distortion, no. When distortion is present the overall reliance of the known to perform the comparison does become more important. QA measures can deal with this. You don't need to mark every single thing in orange just because you saw it in the known first. For example, if I have a 10P quality latent my analysis is a quick "yep it's of value and can be run through AFIS". I'm not spending time marking every feature just for the sake of saying I saw the features before I got a known. If I'm dealing with a really crappy latent, I will do an initial markup of what features are clear and which features are not. I don't use GYRO, instead I mark clear features in pink and features that require interpretation in orange. During a comparison I may end up marking something I marked as pink up front as orange if there is something in that feature that required interpretation during the comparison (maybe the feature is actually shifted but it wasn't apparent in the analysis).

If you need the orange minutiae to tip you over the threshold for sufficiency for an ID are you biased?
I think it would depend of the clarity of the latent. For the comparison in this situation, I would say the reliance on the known to tip you over the edge to an ID could be biasing. It starts to err in the "making it fit" category that makes the hairs on my neck stand up. I agree with Pat that the weight of a feature marked in orange will vary as listed in the instructions so it would depend.

If you consume all the data in order to ID the print, does it diminish the weight of the conclusion?
I'm gonna quote Pat on this one - show me the print.

Should objectivity be a standard for a conclusion?
Our conclusions on what a comparison means is subjective. I don't think you can have an "objective" conclusion since our conclusions are based on our interpretations of data we've observed. Like Pat said, subjective doesn't mean wrong and we need to get out of the mindset of getting defensive when someone says our conclusions are subjective.

If you have to alter your original GYR minutiae from the analysis, should that really cause them to be orange?
Per the instructions, I would say yes. If you had to shift your original mark up around during the comparison, that shifting was due to something in the known being spatially off from what you originally marked. Therefore you're relying on the known for your markup. Since I don't use GYRO in casework, in those situations I would just mark the feature as ambiguous/required some level of interpretation which just so happens to be orange.

Is GYRO a sufficient tool given that it doesn't mark sequence, distortion, or differences in data between the prints?
It's a tool that can be used but it's not the only way to mark up a comparison so I just wouldn't taut it as the end all be all for documentation.

Given that the comments in the analysis were remarkably consistent regarding the presence of overlapping ridges, can an ID be aggregated from two fragments of what is arguably the same print?
If you're dealing with potentially separate touches they should be compared as separate impressions. Especially when significant distortion is present.

If people render a conclusion outside of the majority, should that be considered an error? Should they be held accountable?
First, I'd say since we don't have industry standards on what to rely on for the comparison and how much reliance to put on the comparison you can't really say someone committed an error. You could say their conclusion is inconsistent and may not be justified but since we don't know ground truth in casework we can't really say something is in error. I do think if someone is overstating a conclusion, relying on iffy features to make an ID, etc. there should be accountability. If an agency has specific criteria on when a particular conclusion should be reached in their SOP and an examiner is outside of that then they should be held accountable for being outside of their SOP. But I'd also want to see the agencies SOP to see what their criteria are.

Does your agency utilize a process similar to this where you render a consensus conclusion?
Yes, all AFIS Inconclusives go through a consensus judgement process. This was decided as an administrative policy after communicating with attorneys and investigators for the need for additional QA measures in these situations and clear statements in the report that the weight is extremely limited and does not mean same source. Consensus judgement is also the 3rd step in our Conflict Resolution SOP for disagreements among examiners.

[Boyd Baumgartner]

What if I told you that someone was decertified over this comparison.... What would you say?

[Pat A. Wertheim]

You're a sneaky rascal, aren't you?
Curve ball?
Plot twist?
Ouch!!!

Next question -- are you going to turn over the names of the people who agreed with "identification?"
Ai yi yi!!!