« The Continuing Monkey Wars | Main | The Perils of Profiling »

August 24, 2006

What's wrong with "near-match" DNA database searches?

This week, USA Today reported that the FBI will perform “near-match searching” of CODIS, the Convicted Offender DNA Index System. One might wonder why would police want to know the names of individuals whose DNA profiles fail to match those from samples from crime scenes or rape victims? The answer is simple genetics -- a near match implies that a close relative might be the source of the crime-scene DNA. Police in the United Kingdom have scored several significant successes with the technique. In the words of one official there, "We're having a positive effect on cases (and) preventing additional victims.  What's wrong there?"

Plenty, says Tania Simoncelli, who studies DNA database issues for the ACLU. In her view, “It's the worst kind of privacy intrusion.” Really? Worse than breaking into a suspect’s home and ransacking it for evidence?  Worse than installing a hidden video camera in a bedroom?  Worse than reading my mail or eavesdropping on my telephone calls?  Worse than breaking into a psychiatrist’s office to obtain a doctor’s files and notes?

OK, maybe there are worse invasions of personal privacy, and the ACLU is just engaging in the equivalent of campaign rhetoric here. The question is still what, exactly, is wrong with “near-match searching”? What valid privacy interest can a brother of a convicted offender assert to stop the police from using lawfully acquired information to develop a lead that points to him as a possible culprit? Surely the police can show a mugshot of the convicted offender to a woman who has been raped. If the woman reports that the man who raped her looked a lot like the photo, what is wrong with investigating to see if the brother, who closely resembles the mugshot, might be the rapist? Or suppose that the police question a suspect, who says, “I didn’t do it, my no-good brother did.” The brother might well be upset when the police come knocking on his door, but it is hard to claim that the intrusion on his “privacy” is unwarranted.

What is different when the investigative lead comes from the brother's DNA rather than his face or his mind? --DHK

August 24, 2006 | Permalink


TrackBack URL for this entry:

Listed below are links to weblogs that reference What's wrong with "near-match" DNA database searches?:


"What is different when the investigative lead comes from the brother's DNA rather than his face or his mind?"

Ummmm. One is voluntary and one is involuntary -- the brother chose to implicate his sibling? One is testimony from knowledge and the other is not?

So, now because my brother is no good and was convicted on a crime *I* become a suspect anytime one of my cousins does something illegal?

It sounds like a real misuse of technology to me.

Posted by: ruidh | Aug 26, 2006 5:35:45 AM

I do not think voluntariness is a viable distinction. In the mugshot hypothetical, the brother never consented to having his picture taken. There was no voluntariness there.

In the statement-to-the-police hypothetical, there was voluntariness, but suppose the brother refused to answer any questions when the police interviewed him. Now the district attorney hauls him before the grand jury and grants him immunity. He then implicates his brother. The information pointing to the brother, just like the DNA sample, was not given voluntarily.

The "testimony from knowledge" distinction does not seem to work either. In the mugshot and statement cases, knowledge comes from the perception and memory of the witnesses. In the partial-DNA-search case, knowledge comes from Mendelian genetics, which could be more reliable than testimony from a human observer.

This leaves the argument that you would come to the attention of the police whenever your cousins deposit their DNA in the wrong place. That won't happen very often, and when it does, the inquiry is not likely to go far. It won't happen that often because cousins probably do not share enough alleles to be flagged in a DNA database search for close relatives. Even full siblings only share a bit more than a 1/4 of their alleles. But maybe I am too pessimistic about the ability of the system to distinguish cousins from the general population.

Even if the system routinely led police to cousins, you would have to have at least two of them involved in crimes. One has to have his DNA on file as a convicted offender (or perhaps an arrestee). That sample does not match the crime-scene sample, but it has enough alleles in common to be flagged as a near match. By hypothesis, your DNA was not at crime scene -- you are a false lead. Hence, you must have a second relative whose DNA is not on file, who leaves his DNA in the wrong places and never gets caught and convicted.

Even when this happens, you should be cleared by the DNA test on you if it was your close relative's DNA and not yours found at the crime scene -- the probability of a full 13-locus STR match in two ordinary brothers (not identical twins), for instance, is negligble.

Of course, the issue would not even arise if we had a comprehensive national DNA database. But that is a much longer story ...

Posted by: DH Kaye | Aug 26, 2006 6:27:16 PM

For the sake of clarification, I did not say, nor do I believe that familial searching is "the worst kind of privacy intrusion." What I actually said was that it was "the worst kind of guilt-by-association." I agree that there are far worse privacy intrusions.

The main concern I have with familial searching is that, if routinely practiced, it would effectively place hundreds of thousands of innocent people who happen to be relatives of persons convicted of a crime under lifelong genetic surveillance.

Posted by: Tania Simoncelli | Sep 28, 2006 8:19:05 AM

The _USA Today_ article quoted you as saying that "It's the worst kind of privacy intrusion," so we appreciate the correction. Whether it is "the worst kind of guilt-by-association" is another question, and I am a bit skeptical of that claim as well.

It certainly is true that the practice exposes very close relatives not in the database -- both guilty and innocent -- to at least temporary suspicion. At the same time, the "genetic surveillance" here does not consist of deciphering the genome, and if this is "guilt by association," then so is any inquiry that leads investigators to question individuals who are associated with other suspects. A worse form of "guilt by association," I would have thought, occurred in the days of Joseph McCarthy, when the ACLU valiantly defended the targets of such denunciation.

Obviously, this is not an argument for near-match searching, but I am still looking a convincing distinction between this investigative practice and more established ones that can lead police to look into the conduct of (in the aggregate) hundreds of thousands of innocent people. Generating suspects and clearing most of them is how police often work, so the issue I am curious about is why DNA-type leads are worse than phenotypical ones.

Posted by: David Kaye | Nov 11, 2006 6:27:59 PM

Post a comment