April 17, 2017
UPDATE: May 3rd, 2017 — New information to this case at bottom of article.
Idaho Falls police need to stop watching sci-fi crime movies and pseudo-scientific crime scene investigation TV shows.
I was watching ‘48 Hours: The DNA of a Killer‘ a couple of nights ago (aired April 15th but I DVR’d it). I was shocked at the amount of bad ‘science’ and poor investigation practices being employed by the Idaho Falls Police Department.
Almost no evidence, twenty-three days of interrogations, seven highly manipulated polygraph sessions, and overall lackluster police work were all mechanisms of this nightmarish perversion of justice. The finale of this department’s atrocious understanding of science was a claim they made near the end of the episode. They had a DNA facial sketch — a 3D computer image generated from the DNA sample — and they intended to use it to find the murderer.
Although the idea of developing a physical profile from DNA is not completely stupid, creating a facial recognition quality image has not held up to any real scientific scrutiny; but more on that later.
The story starts in Idaho Falls, Idaho after 18 year old Angie Dodge is stabbed to death. A very unreliable acquaintance of Angie is implicated as an accomplice to the crime and confesses to holding Angie during the stabbing, although the confession is very questionable and is later recanted. There is ‘male evidence’ from the killer found at the scene but no other evidence. That should be excellent evidence right? Well, only if they can match it to someone. The problem is it doesn’t match the alleged accomplice and they can not find a match in any law enforcement database. So they make the highly dubious decision to turn to ancestry.com and while they don’t find a match to the killer, they do find a possible familial match. The killer might be a relative of an ancestry.com customer.
This, in my eyes, opens up a gigantic can of worms involving the privacy concerns of examining a database of millions of innocent people, the uncertainty of collection and testing procedures, the question of how close a familial match is sufficient to be useful, and the evidentiary chain of custody issues. However; as I watched this case unfold, I was really taken aback by the claim of getting a facial ID sketch from the DNA sample.
For now, I will focus on that idea of DNA generated facial recognition. I’ll stay away from the issue of the completely useless polygraph they employed since I have already expressed my opinions on those pieces of junk.
I understand the police are very motivated to find a killer and frustrated by their inability to do so, but belief in techniques which aren’t shown to be real has some very real and dangerous permanent consequences.
Which leads us back into the story.
The investigators manipulate and possibly coerce an apparently false confession from the ‘accomplice’ who gives the police several different murderer suspect’s names during the multiple interrogations.
One of the names given is Michael…
The search of the ancestry.com database turns up a familial match to a man who has a son named Michael, an independent horror filmmaker from New Orleans. Michael now becomes their prime suspect. Their ‘evidence’ to implicate this suspect is the controversial and vague familial DNA match to his father from ancestry.com and the fact that his name is Michael, one of the top five most common American first names. On top of that, Michael makes creepy horror films — So they’re certain they have the guilty man.
This logical leap from a vague, slightly adjacent DNA match to suspect was so large that I was getting confused even typing that sentence.
Top five most common first name, and a horror film maker… we might want to lock up John Carpenter next.
That’s some mighty fine police work there, Lou!
It turns out that Michael is innocent and has never met Angie — no surprise here — but the investigators now say they have a facial ID sketch that they generated from the DNA sample of the murderer and will use it to track down a new suspect. Sigh… They haven’t figured out anything based on the evidence they actually have so now they will focus on unproven technology.
The DNA snapshot or ‘DNA phenotyping‘ seems to just be profiling based on ancestral region with a strong racial component. Based on a person’s DNA, you might be able to make some very broad generalizations about facial structure; cheekbones, brow, and nasal area. You might also be able to fairly accurately determine skin, eye, and hair color. Probably a good test for determining race but not much more. I don’t think DNA alone is nearly enough to make a reliable identification of a person’s overall facial appearance.
There have been several instances of these facial images being released to law enforcement to help find suspects (a couple of cases even released publicly) but I was only able to find one instance where a DNA composite image might have uncovered new leads because it correctly identified race.
In that case the eyewitnesses and FBI profiling indicated the perpetrator was a Caucasian male. Investigators tested DNA from thousands of Caucasian males and found no match. DNA phenotype testing indicated the ancestry of the suspect was 85% sub-Saharan African and that changed the direction of the investigation. Police then arrested Derrick Todd Lee, who was later convicted for two murders.
The police already had confined their search from the start to a Caucasian which was blinding them so this test opened them up to look at other possibilities which is good but to give credit for solving this crime to just this technique seems incomplete and unscientific.
One use of this technology that seems more likely to produce a successful result was a case of attempting to ID a body that was found with the skull intact. A DNA composite image was created by Parabon NanoLabs using the victim’s DNA. Now that might be promising. The victim’s skull being present would take some of the uncertainty away. It might be a useful technique to help identify Jane/John Doe bodies, especially if it can then be matched to a missing person’s list of photos.
This technology might have a future but it is my opinion that it is not ready to be completely relied on for real police work yet, and it certainly won’t work as a facial ID sketch the way the Idaho Falls police were claiming on this ’48 Hours’ episode.
I wanted to post example images but couldn’t find any that wouldn’t risk copyright claims so you can see for yourself here on their own site – Parabon’s Snapshot™ DNA Phenotyping Service.
[UPDATE 5/3/17: I can now post the image they released today]
They make much more optimistic claims than I have seen in real-world examples but they do have several examples on their website they say proves how well it works. Sidenote: If they are trying to promote this cutting edge technology, why does their website look a little dated? — Contact me, Parabon… my rates are reasonable 😉
Anyway, there were many things wrong with this case I just wanted to point out a couple of the magic-CSI stuff that always sort of rubs me the wrong way. Science is awesome and we can find out a lot with science but bad/pseudo-science is very dangerous and can affect innocent people’s lives.
UPDATE, May 3rd, 2017:
I just saw this today… see news article below. The Idaho Falls Police have released the phenotype facial snapshot. This is a very interesting development in the time since I wrote this blog article. They are very interested in solving Angie’s murder as is apparent by the money and effort that has been put into this case. I will be watching the outcome of this case very closely. I am still extremely skeptical of this technique. The image looks like a fairly generic twenty-something white man. I am hopeful for a resolution to this case and I will be fascinated to find out if any real leads come out of releasing this image and profile. I have a suspicion however, that it will probably generate a lot of people calling in saying, “Hey, that guy looks like my neighbor… also he kinda looks like a guy I work with… or maybe the mail man…” — CSH