Friday, January 18, 2013

CODIS Interruptus

The law enforcement community (police, DAs, and the courts) have figured out that they can usually avoid unwanted complications from engaging in CODIS if they begin but do not complete the act.

CODIS stands for the Combined DNA Index System. The term, as loosely applied, incorporates also the NDIS, the National Index System. It's more difficult to make a clever title from NDIS, so I'll use the acronym CODIS to encapsulate both systems, and a bit more.

From the FBI website we learn the following:
CODIS—NDIS Statistics
Measuring Success
The National DNA Index (NDIS) contains over 10,086,400 offender1 profiles, 1,332,700 arrestee profiles and 466,800 forensic profiles as of December 2012. Ultimately, the success of the CODIS program will be measured by the crimes it helps to solve. CODIS’s primary metric, the “Investigation Aided,” tracks the number of criminal investigations where CODIS has added value to the investigative process. As of December 2012, CODIS has produced over 198,400 hits assisting in more than 190,500 investigations.
There are more than 11 million possible matches for any DNA sample sent to CODIS. That's impressive, to be sure. Still, that is only one-tenth of 100 million individuals whose fingerprints are held in AFIS (Automated Fingerprint Identification System).

As a database programmer, I'm confident that searching for a 12-allele DNA match (among 10 million records) is not fundamentally more difficult than searching for a 10-point fingerprint match (from 100 million individuals, each having approximately ten fingers).

Despite the frequency of CODIS and AFIS comparisons, I have found it surprisingly difficult to determine the unit search cost for either CODIS or AFIS. This (excellent) article suggests that the marginal cost of a DNA database profile is less than $40. From this site, we can infer that the cost of a fingerprint search of AFIS is around $15. The point is that the marginal cost of a single CODIS or AFIS search is tens of dollars.

As technology improves, the unit search cost will drop to dollars, then to pennies. Already Google, Bing, and other search engine operators will, for zero pennies, compare any arbitrary string of text you can come up with against the entire internet, and sort millions of possible hits according to those they believe are most appropriate.

While a database search is (or should be) inexpensive, generating a numerical DNA profile from a biological sample still takes man-hours. If you want to have your DNA profiled by a private firm, it seems as if it's going to cost you between one and two thousand dollars. Once you have that profile, however, you could submit it to CODIS as many times as you wish, assuming you were allowed to access CODIS. The first comparison would cost a few thousand dollars, including the cost of the profile, but subsequent runs would cost just tens of dollars.

But why, you might ask, would you spend even a few tens of dollars if you've already attempted to find a match in CODIS and came up empty? The answer is simple, assuming you are not a district attorney or an appellate court judge.

New DNA profiles are being added to CODIS as we speak. I present one chart from this (excellent) presentation.


Wow!

Now let's consider all this in light of the impending execution of Larry Ray Swearingen, The Most Innocent Man on Death Row. In that case, the State found blood flakes underneath one fingernail of victim Melissa Trotter. They extracted a DNA profile from those flakes and compared that profile against the DNA profiles of Melissa Trotter herself, Larry Swearingen (who is scheduled to be executed for her murder on 27 February), and Robbie Shore (who had allegedly, repeat allegedly, been harassing Melissa Trotter.) All three were excluded as the contributor of those blood flakes.

On that evidence alone, Larry Swearingen should probably have been set free unless the State had equally compelling evidence of his guilt. They did not, so he should have been set free. We're far beyond that argument, however, and I need to focus.

In my recent series Who Killed Melissa Trotter, I argue that Anthony Allen Shore is a far more likely culprit than is Larry Swearingen. (As it turns out, anyone who hadn't been in jail during the two weeks before her murder is a more likely candidate than Larry Swearingen.) Shore confesses to strangling five young women to death and that he attempted to strangle a sixth as he raped her. He was finally identified by a cold case DNA submission to CODIS in 2003. It's not clear how long his DNA profile had been in CODIS before that hit.

As part of a plea bargain for child molestation charges, Shore gave a DNA sample in 1998. CODIS was still in its infancy, and the Houston Crime Lab was still too crappy and too corrupt to process the DNA rapidly. I'll present just one brief snippet from the 300-page report of the independent investigator hired by the City of Houston and the HPD to look into press reports of problems at the crime lab.
[T]he Lab’s historical serology and DNA work is, as a whole, extremely troubling. We found significant and pervasive problems with the analysis and reporting of results in a large proportion of serology and DNA cases. The Crime Lab’s substandard, unreliable serology and DNA work is all the more alarming in light of the fact that it is typically performed in the most serious cases, such as homicides and sexual assaults. On the whole, this work did not meet the generally accepted forensic science principles that existed at the time and posed major risks of contributing to miscarriages of justice in extremely significant cases, including death penalty cases.
Okay, two snippets.
Although, as in most crime laboratories, the vast majority of cases referred to the Crime Lab involved controlled substances, the demand for DNA analysis increased substantially in the late 1990s and early 2000s. During this period, as discussed in greater detail later in this report, rape kits became substantially backlogged with respect to cases in which there were no known suspects. These cases were not analyzed and therefore not loaded into the Combined DNA Index System (“CODIS”) database. The DNA Section’s de facto policy at the time, developed because of the workload demands placed on its limited personnel, was to conduct DNA analysis only in cases involving known suspects from whom samples had been obtained for comparison. [Footnote: Mr. Krueger told us that, even though DNA cases typically involve serious crimes such as homicides and sexual assaults, controlled substances was his priority for resources because of the large volume of drug cases that were flooding the Crime Lab.]
There you go. It is extremely unlikely that Shore's DNA was even in CODIS when the State profiled the blood flakes found under one of Melissa Trotter's fingernails. The HPD was too busy working on drug cases to perform the work necessary to identify rapists, child molesters, and serial killers. Drugs may or may not be a victimless crime, but the War on Drugs is certainly not.

Even if the State had submitted the blood-flake DNA profile to CODIS, my point (which I am on the verge of getting around to) will be unchanged.

Anthony Shore may have killed Melissa Trotter, and Anthony Shore's DNA may not have been in CODIS even if Texas had bothered to run the blood-flake DNA profile through CODIS. If we knew all that to be true, it would be unconscionable to execute Larry Swearingen without searching for a hit. That is still not quite my point.

My point is that it is unconscionable to execute anyone anymore without running all possible probative DNA through CODIS at least once, recently, as a condition for issuing the death warrant. In Larry Swearingen's case, I don't know that any DNA has ever been run through CODIS. There's plenty of DNA material to be tested, and there are now 11 million possible contributors that were not there at the time of Swearingen's trial.

How can a State, in good conscience insist on executing someone while probative DNA goes untested against the current CODIS population? Even if the DNA had been run through CODIS ten years earlier during the trial, the actual killer's DNA could have been added at any time since.

My understanding is that every Monday, the entire CODIS DNA profile database is compared against the DNA from new crimes. I argue that every Tuesday, the entire CODIS DNA profile database should be compared against the DNA from cold cases, and that every Wednesday the entire CODIS DNA profile database should be compared against every closed case.

As an absolute, bare-bottom, gosh-darn minimum, all probative DNA should be tested against the current CODIS database as a condition of even requesting a death warrant.

That, finally, is my point.

"An idea starts to get interesting when you get scared of taking it to its logical conclusion." --Nassim Nicholas Taleb