CLPEX Chat Board

Does anybody have a complete list of every erroneous identification made in the history of latent print identification? I believe Dr. Simon Cole has spent the last couple of decades trying to compile such a list. To the best of my knowledge last time I checked, he had not yet reached 30. No doubt, of course, there are more that are not to be found in the literature. Let's be generous and give it a factor of 10. So if there have been 300 erroneous identifications responsible for convictions in the last century, how many correct identifications have there been? And considering other factors such as incompetent people doing comparisons (as in the Lana Canen case), how many of those erroneous identifications have been influenced by bias? So what percentage of cases have been erroneous identifications brought about by bias? I would submit that if you say one in a million, you are greatly exaggerating the truth.

I have never seen any evidence that bias is a real problem in any significant number of cases. One case, Brandon Mayfield, may be significant to Mr. Mayfield, but in the century long history of fingerprint identification, his single case cannot be considered a significant percentage of all cases.

Mr. Dror's extreme cases and manipulation of evidence cannot support the idea of bias in real casework.

Sorry. Bias is NOT a problem for ethical casework examiners conscientiously doing their jobs day in and day out. Don't throw the baby out with the bathwater. And please don't tell me Dror's class will make me a better examiner. It will only serve to enhance his ego.

DISCLAIMER: I AM NOT AN EXPERT ON COGNITIVE BIAS!

However, bias, like errors, have different forms. We've read (or heard about) the bias in Mayfield. We've looked at other erroneous identifications and many were faulty reliance on L2D or even L3D. Some erroneous ID's were made by improperly trained examiners. Some were because of a lack of adherence to ACE-V.

Michele Triplett's online fingerprint dictionary lists around 40 erroneous ID's (some were reported in the media) and not all have the causes for the errors. I think what we as examiners need to be cognizant of is how we can minimize the myriad of causes for errors. The reason erroneous ID's are looked at as the worst kind of error is because they have the potential to put someone in jail for a considerable amount of time depriving them of life and liberty.

There is some research being conducted by Dr. Busey of The Indiana University where he's collecting data on what people find are acceptable tradeoffs for errors committed in fingerprint identifications and exclusions. The link to participate in the study is https://goo.gl/JLGyVy I'm excited to see the results of this research (maybe more of an thought experiement for now) and how people view errors. There are some basic demographic questions at the end to see how certain groups of people are responding to the sliding scale question. This might give us a better sense of what's viewed as acceptable (regardless of it's right or wrong).

So, is this class an attempt for Dr. Dror to capitalize on training that the NAS report was saying was necessary? Maybe. Could it be an opportunity to understand bias more and, if what you are saying is true, articulate why bias doesn't pose a problem for casework examiners? Of course. I think it's important to at least know what our critics are saying (and Dr. Dror isn't an opponent of fingerprint comparison and its use in identification) but if, at the end of the day we can say, "Yes, I know there might be bias but I do everything that I can to minimize it (because of x, z, and z)", then I think it could be beneficial. Does that mean I'm going to be the first one to sign up for it? Absolutely not. I don't jump on Kickstarter projects and I don't usually jump on a brand new class until I have heard some reviews about it first.

Either way, I don't think what Dr. Dror is saying is that we are all bad examiners because there is so much bias present and we are failing to prevent it. I think what he is saying is that contextually, it can be there for certain reasons and if we build a process or system to not only address those biases but minimize them, we are doing good work.

Good news, Josh. You don't have to be an expert, just an informed consumer.

We tend to approach bias in one of four ways: 1) check under our proverbial beds for the bias boogeyman because it scares us 2) speak about it in such vague terms, it's essentially meaningless 3) reference it as such a high concept that we can't possibly understand it, but we're certain it exists.....everywhere 4) Put it in a root cause analysis document surrounding some error because.....bias! I find these approaches unsatisfying only because they give bias wholesale acceptance without measure, which is ironic in some sense because that was the critique against fingerprint identification. The categories of bias are so broad that it's possible to construct a narrative that bias influences all and every aspect of an examination and it's therefore inaccurate. But just because it makes sense, doesn't mean it is so, and the data bears that out.

As a practicing examiner and one who has seen Dror in a variety of forums, I find his work to be a little, well......... biased. I tend to see conversations about uncertainty and risk converging around the single ID from AFIS, especially when that comparison ends up being complex and therefore bordering on inconclusive - consistent but insufficient. To those ends, I like the idea that he and Kasey came up with regarding the bias that AFIS ranking can potentially have.

Onto PCAST/black box and its implications. If scientific validity for subjective measures according to PCAST are only established through black box studies, then shouldn't bias studies be subjected to the same standards? The black box study clearly demonstrated the direction and magnitude of error was biased towards wrongful exclusions, so you'd think that'd be an area to easily exploit in a study. One of the things I liked about PCAST was their critique that the forensic disciplines are inbred (Read: Not married to scientific disciplines/journals) and as such, why not learn from other disciplines? It's possible to design experiments to induce cognitive bias. If you can induce bias, in theory you can reduce bias and demonstrate that ability. I'm into science that measures and projects itself onto the task of solving problems, not conjecture that keeps the spectre of problems ever present in order to capitalize from the fear.