How private DNA data led Idaho cops on a wild goose chase
May 3, 2015 1:49 AM   Subscribe

... and linked an innocent man to a 20-year-old murder case. Analysis by the EFF of the case of Michael Usry, a New Orleans filmmaker whose father's DNA profile in a non-profit DNA database, which he had been assured would remain private, dragged him into a grisly unsolved murder case.

This new search for evidence in the murder of Angie Dodge seems to have been spurred by questions about the conviction of another man, Chris Tapp, for the crime. The DNA found at the scene doesn't match him either, and there may have been police misconduct in the investigation.

---

We've talked about familial DNA profiling before, when it was used to find the Grim Sleeper serial killer and, controversially, the killer of Yara Gambirasio. We've also discussed the Phantom of Heilbronn, the mysterious one-woman crime spree who turned out to be a worker at a factory making cotton swabs for evidence collection.
posted by daisyk (40 comments total) 19 users marked this as a favorite
 
Whoa, that's scary.
posted by Joe in Australia at 2:43 AM on May 3, 2015 [1 favorite]


Ok, if cops are going to use DNA matching like this, we've got to find a better way to explain the statistical fact that running lots of tests decreases the meaning of getting a hit on any single test.

A random match of 34/35 allelles is statistically unlikely; however, the chance of finding such a meaningless match when you check against a very large database is *not* necessarily unlikely.

It's the Birthday Problem (only with larger numbers and a less-random underlying data set), and I swear, any forensic DNA expert should have to explain that problem correctly before they're allowed to practice.

And this doesn't get into the issues about privacy or familial searching in general.
posted by nat at 2:44 AM on May 3, 2015 [40 favorites]


Not just practice. Any forensic DNA expert should have to explain it correctly in court before she/he can offer any other testimony.
posted by rmd1023 at 3:59 AM on May 3, 2015 [4 favorites]


Court's too late, though. Michael Usry ended up ok, here, but plenty of other people who were just being investigated haven't.
posted by nat at 4:40 AM on May 3, 2015 [1 favorite]


Makes me thankful I'm a Kiwi.
posted by Samuel Farrow at 5:14 AM on May 3, 2015


Being hexaploid must mean that you never get fingered for any crimes committed by humans.
posted by XMLicious at 5:31 AM on May 3, 2015 [15 favorites]


Not just practice. Any forensic DNA expert should have to explain it correctly in court before she/he can offer any other testimony.

Nice thing about this is no regulator has to adopt the rule. Any defense attorney should be able to challenge any expert on this I think. In fact you could probably write up a multiple-comparisons script that could be used in a lot of cases, if these practices are really wide spread.
posted by grobstein at 5:32 AM on May 3, 2015 [1 favorite]


Being hexaploid must mean that you never get fingered for any crimes committed by humans.

Keywords: Hexaploidy as the key insight.
posted by grobstein at 5:33 AM on May 3, 2015 [2 favorites]


Yeah, the real key is that you have to find ways to discount this kind of testimony for juries without scientific training and with a serious cultural deference to forensic evidence from decades of magic CSI shows.

I don't think we should be very happy if that involves turning the court room into an intro stats class. Juries aren't as smart as the average college class and lots of college students fail stats.
posted by anotherpanacea at 5:45 AM on May 3, 2015 [12 favorites]


Since this guy was never tried (or even arrested AFAICT), he didn't have an opportunity to question the police's DNA strategy. And they only used the familial test in their warrant application to get his individual DNA, which would have allowed a match with much higher confidence if he was the attacker.

So both not a lot of harm done, and not a lot of upside to challenging the familial match in court, I suspect. "34/35 alleles" match could be enough for probable cause even if there is some probability of such a match with no familial connection.
posted by grobstein at 5:49 AM on May 3, 2015


Also, I would like to see a lawsuit against the private DNA database that comes down so hard it starts putting the fear into companies that would break their data agreements so cavalierly.
posted by rmd1023 at 5:57 AM on May 3, 2015 [28 favorites]


Yeah, the real key is that you have to find ways to discount this kind of testimony for juries without scientific training and with a serious cultural deference to forensic evidence from decades of magic CSI shows.

I don't think we should be very happy if that involves turning the court room into an intro stats class. Juries aren't as smart as the average college class and lots of college students fail stats.


Yeah sort of a tough issue, but there should be ways you can present the question so as to minimize the base rate fallacy. There's some research in that area.

It also might be a question for the judge.
posted by grobstein at 5:57 AM on May 3, 2015


Also, I would like to see a lawsuit against the private DNA database that comes down so hard it starts putting the fear into companies that would break their data agreements so cavalierly.


I had this thought also! It was supposed to be private!

Unfortunately(?), I'm not sure there are meaningful damages in this case.
posted by grobstein at 5:58 AM on May 3, 2015 [1 favorite]


Also, I would like to see a lawsuit against the private DNA database that comes down so hard it starts putting the fear into companies that would break their data agreements so cavalierly.

Personally, I would like a DNA database to cryptographically blind my identity, so they and their unknown corporate successors can't share it against my will. Obviously no legal privacy agreement is strong enough for this.
posted by grobstein at 6:01 AM on May 3, 2015 [7 favorites]


And this is why when my own DNA was tested (long boring story), I gave the testing company a fake name, and no other specific ID info.
posted by easily confused at 6:44 AM on May 3, 2015


The important thing here is that this was done without a warrant -- the private db just gave the police what they wanted when asked.
posted by Ivan Fyodorovich at 6:48 AM on May 3, 2015 [13 favorites]


"It also might be a question for the judge."

But how many judges have any statistical understanding? Like hair and bite analysis, DNA can be a dog and pony show. Expert witnesses for the prosecution who want to work more cases know exactly what they have to say to make their money.
posted by idiopath at 6:56 AM on May 3, 2015 [2 favorites]


I'm deeply disturbed that Sorenson, which runs a private forensic lab based in Utah, would sell its DNA data to Ancestry.com. Even if it had been completely blinded and provided only for "if you have these markers your ancestors might have come from $COUNTRY" population genetics reasons, that would still seem highly unethical.
posted by Flannery Culp at 6:58 AM on May 3, 2015


Expert witnesses for the prosecution who want to work more cases know exactly what they have to say to make their money.

No, we know exactly what we have to say in order to be scientifically correct yet also understandable to jurors who may have last thought about DNA and genetics in 10th grade biology. It's a difficult balance to strike and further circumscribed by the adversarial court system where you can only answer the question asked, not what you think the lawyer meant to ask or maybe should have asked.

I am always very upfront that no, I cannot tell you how that DNA got there, and no, I cannot conclusively say it was his even if they match at every allele, and yes, we can talk about secondary and tertiary transfer if you like. Any forensic scientist worth the name should be able to acknowledge the assumptions and limitations inherent in their testing methods.
posted by Flannery Culp at 7:04 AM on May 3, 2015 [7 favorites]


I can tell people whether one of my great-grandparents committed adultery; and I can tell people whether members of my family will have a marker associated with mental illness, even if they haven't been born yet. I can tell people exactly which one of my ancestors crossed a color barrier, and where my family tree might have been a little too straight of a line. I don't even have to know any of this myself; and I don't even have to know the future people on whose behalf I'm also telling it; and I can't possibly know what the future context of all this information will be. All I have to do is spit in a cup.
posted by Hypatia at 7:34 AM on May 3, 2015


In the UK at least judges have refused to allow basic statistical information about the base rate fallacy to be admitted as evidence.

I regularly teach the case of a woman who had two SIDS deaths: prosecutors argued that this was astronomically unlikely and so she must have murdered the children. In fact, there's about a 2/3 chance she didn't. (And she was eventually released after being convicted and spending three years in jail.)
posted by anotherpanacea at 8:06 AM on May 3, 2015 [25 favorites]


Wow, thank you for sharing this an-pan! Scary stuff. I wonder how the issue has been treated in US courts.

(Even if you can't talk about the base rate fallacy, perhaps you can phrase your statistics in ways that are less susceptible to the base rate fallacy?)
posted by grobstein at 8:22 AM on May 3, 2015


In the U.S., no prosecutor would ever undermine her case by avoiding possible statistical fallacies. (Look at the recent challenges to absurdly prejudicial PowerPoint slides.) Taking advantage of base rate errors is just "being persuasive." You'd have to get a major set of appellate actions to discourage statistical chicanery, and even then....
posted by anotherpanacea at 8:31 AM on May 3, 2015 [3 favorites]


Given that statistics are hard, and easy to abuse, why not rule out statistical arguments altogether?
posted by idiopath at 8:42 AM on May 3, 2015 [1 favorite]


Idiopath: because fundamentally very little in a trial is 100% undeniable and therefore probability and statistics are really all you have?
posted by edd at 8:51 AM on May 3, 2015 [2 favorites]


> Idiopath: because fundamentally very little in a trial is 100% undeniable and therefore probability and statistics are really all you have?

Then the jury needs to be accurately educated on statistics. Allowing statistics to be used as a magic weapon where the jurors just nod their heads credulously and then vote to convict is a very bad idea.
posted by lupus_yonderboy at 8:54 AM on May 3, 2015 [1 favorite]


The reasoning behind this stuff becomes much funnier if you swap out the sciency for things people actually experience.

"The likelihood of wearing this hat is less than 1 in a million. The murderer wore this hat, and so does this man here."
posted by idiopath at 9:01 AM on May 3, 2015 [4 favorites]


Lupus_yonderboy: I'd love that, but if you're going to make understanding statistics a condition of being a jury member you're going to have a very hard job finding enough jurors.

Judges however - you should surely be able to train them well enough?
posted by edd at 9:02 AM on May 3, 2015


Judges however - you should surely be able to train them well enough?

If you're not so stupid as to have a system that depends on elected judges where knowing nothing is an advantage to getting elected.
posted by MartinWisse at 9:04 AM on May 3, 2015 [10 favorites]


anotherpanacea, we had a case like that here in the Netherlands.
posted by Pendragon at 9:20 AM on May 3, 2015 [3 favorites]


This is when the case starts to sound like something out of the TV show “CSI.” Sorenson linked the crime scene DNA to DNA from a man born in 1952. That man didn’t fit the age profile of the murderer, so the cops used Sorenson’s genealogical information to trace his male descendant line and find his son, Michael Usry Jr., born in 1979. Then the cops searched Usry’s Facebook page and found he had some Facebook friends who lived somewhat near Idaho Falls. And then through Google searches, the cops learned Usry was a filmmaker who had been involved in making a few short films that had homicide or killings in the story line. (The cop noted in a warrant affidavit “these short films have won awards in several film festivals.”) Based on this completely circumstantial evidence, the Idaho investigators got a warrant to collect a swab of Usry’s DNA.

Good lord. That isn't even "circumstantial evidence" -- it's just straight-up coincidence. This is just scary and dovetails nicely (and unfortunately) with all of the other scary shit "law enforcement" is doing these days.
posted by mudpuppie at 9:57 AM on May 3, 2015 [5 favorites]


It isn't the Birthday Problem though. That would be finding a two sets of matching alleles where either set is free to range over the complete population. The particulars of how the matching was done, and the probabilities of the various single alleles - could have a very big difference on the result.
posted by Death and Gravity at 10:27 AM on May 3, 2015


Then the jury needs to be accurately educated on statistics.

What kind?
posted by MisantropicPainforest at 10:30 AM on May 3, 2015


Then the jury needs to be accurately educated on statistics.

if you're going to make understanding statistics a condition of being a jury member you're going to have a very hard job finding enough jurors.


Defense attorneys and prosecutors routinely dismiss jurors who display specialty knowledge or education. The problem isn't finding people smart enough to understand statistics. The problem prejudice against them.
posted by clarknova at 11:16 AM on May 3, 2015 [4 favorites]


Good lord. That isn't even "circumstantial evidence" -- it's just straight-up coincidence. This is just scary and dovetails nicely (and unfortunately) with all of the other scary shit "law enforcement" is doing these days.

It's not just police (although this kind of crap is the rule not the exception). Judges have to help them in these bad faith efforts.


I know someone who was denied a Miranda motion because, while the arresting officer testified explicitly that he was detained without being read his rights, the judge sustained an objection against that statement, and went on to offer that the cop was too short to be intimidating. Ergo the guy had no reason to believe the officers in his bedroom were detaining him.

It requires law enforcement, prosecutors, and judges working in harmony to railroad people like this. The Innocence Project found that 1/4th of all death row inmates were exonerated by DNA evidence when it became available. That's a 25% false conviction rate in cases that require the highest standard of evidence. Imagine how bad it must be for lesser felonies and misdemeanors.
posted by clarknova at 12:06 PM on May 3, 2015 [9 favorites]


Oh my goodness! My DNA is in that database. Sorenson was a Morman who made a lot of money in the medical device business. I assume he set this up as a way for the Morman church to get more DNA for the whole Morman 'convert the dead' thing. I have no problem with that and they were offering DNA analysis for very low cost--much less than anyone else. I was using it for family ancestry research.

So both not a lot of harm done, and not a lot of upside to challenging the familial match in court, I suspect. "34/35 alleles" match could be enough for probable cause even if there is some probability of such a match with no familial connection.--grobstein

The thing is, that is completely false. If this was 35 random numbers, then a match of 34 would be a very strong indicator of close relationship. But it isn't. Large groups of us share a lot of these alleles. I'd be willing to bet that 90% of Mefites match 25 or more alleles. It doesn't take much to make that number go higher. In the Sorenson database alone, my DNA had a match of 34/35 for two different families, with no known relatives going back several generations! There were a lot more matches for 32/35.

So these police had no idea what they were doing. Now I'm worried. Are my kids going to be brought in for questioning because someone in one of these 34/35 families commits a crime?
posted by eye of newt at 12:10 PM on May 3, 2015 [6 favorites]


My point was just that, since probable cause is a lower standard of evidence than beyond a reasonable doubt, a court might agree that the familial match gives rise to probable cause even if it's the kind of test that would likely generate a few false positives on the data. The police would also argue that the familial match was further supported by Usry's connections to Idaho.
posted by grobstein at 12:15 PM on May 3, 2015 [1 favorite]


I'm speculating about the law here -- I just wanted to clarify I wasn't saying that the familial match was anything like conclusive evidence.
posted by grobstein at 12:17 PM on May 3, 2015


The problem in the FPP is just the tip of the iceberg: lots of populations have smallish gene pools, and big DNA databases will inevitably find matches among them. Ashkenazi Jews, for instance, get lots of bogus matches at the second- or third-cousin level on the 23andMe database because they're descended from a small founding population - and that's with only a fraction of a percent recorded. The problem will get greater as the databases expand.
posted by Joe in Australia at 6:50 PM on May 3, 2015 [3 favorites]


Could we assign probabilities to the various legal standards? Like "beyond a reasonable doubt" is 95%? Probable cause is 50%? Reasonable suspicion is 20%, etc?

In tort law the "preponderance of the evidence" standard maps pretty well to 50%, which is why I ask. I'm under the impression that US courts have actively resisted this kind of statistical evaluation, and that this leads to some interesting possible Dutch Book problems.
posted by anotherpanacea at 11:32 AM on May 4, 2015


« Older Shit people say to Women Directors   |   I'll help you find your own reason Newer »


This thread has been archived and is closed to new comments