| Sun | Mon | Tue | Wed | Thu | Fri | Sat |
|---|---|---|---|---|---|---|
| 1 | 2 | |||||
| 3 | 4 |
5 |
6 |
7 | 8 |
9 |
| 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| 17 | 18 | 19 |
20 | 21 | 22 | 23 |
24 |
25 |
26 |
27 | 28 |
29 |
« NSA data collections | Main | Citizen Relationship Management: the rocky road from transactions to empowernment »
11 May 2006
On a related note, regarding the data mining of relational data, today with co-authors Frederick Bieber and Charles Brenner, I had a paper come out in Science on using DNA databases to identify relatives of those in the databases: "Finding Criminals Through DNA of Their Relatives." In the extended entry is the Harvard press release.
In a small nutshell, we use a combination of Monte Carlo simulations and genetic kinship analysis to demonstrate the feasibility of identifying the source of a crime scene sample if a close relative of the source is in one of the offender DNA databases maintained by various jurisdictions around the country (and the world). We find, for example, that in a database of 50,000, if one rank ordered the most likely close relatives to least likely close relatives, if a close relative were in the database, that relative would be ranked first about half of the time, and in the top 100 99% or so of the time. Accuracy could be improved further if additional genetic data were extracted from the samples. In short, kinship analysis would indirectly incorporate most of the first degree relatives of those in DNA databases (there are 3+m people in US databases). We discuss the policy and ethical implications of these findings, which I will not go into here (see release below).
There are two interesting things, from the perspective of this bog, that I would like to note. First, is the paradigm shift that kinship analysis represents. If one looks at the discourse around the expansion of DNA databases in this country, it has focused on the attributes of those going into the database-- convicts for certain crimes, what is the probability of recidivism, etc. But genetic data is inherently relational, and thus the technology is really incompatible with the discourse. The interesting thing--although, perhaps this paper will just drop into the bottomless well, never to be heard of again-- is what happens when credible research comes out highlighting the disjuncture.
Second, is the more general issue around data mining of relational data, individual choice, and privacy. As both this article and the NSA data mining highlight, so much information is inherently relational. The fact that I have certain characteristics may say something about people that I have various types of connections to. This does create certain conundrums for constructions of privacy that rely on individual choice, since with relational information, what I choose to reveal about myself reveals something about others-- i.e., there are informational externalities. In this day and age, this issue is endemic, and suggests that certain types of decisions about privacy must be necessarily communal and not individual in nature.
Using Relatives' DNA to Identify Criminals
Potential problems and benefits identified in new research study
Genetic analysis of proliferating DNA databases maintained by the criminal justice system might enable criminals to be caught by recognizing, as their kin, those in these databases. While use of these novel methods could provide thousands of new investigative leads, they also raise challenging legal and policy questions. These findings along with discussion of the legal and policy issues are published in ScienceExpress, the advance online version of Science, on May 11, 2006.
The article is co-authored by Frederick R. Bieber, PhD, a medical geneticist at Brigham and Women’s Hospital and Associate Professor of Pathology at Harvard Medical School; Charles H. Brenner, PhD, a forensic mathematician and Visiting Scholar at the School of Public Health, University of California, Berkeley; and David Lazer, PhD, Director of the Program on Networked Governance and Associate Professor at Harvard’s Kennedy School of Government.
“We demonstrate that the kinship analysis methods routinely used for humanitarian mass disaster and missing persons identifications could also be used to identify criminals who are close relatives of those in the rapidly expanding offender DNA databases,” says Bieber.
Currently there are about 3 million individuals in offender databases around the country, largely including individuals convicted and arrested of serious crimes. The genetic profiles of these individuals are then compared to evidence of crime scene evidence, potentially linking known individuals to crime scenes. The team evaluated the potential for using these databases to link crimes to individuals related to those in the database by analyzing data on existing genetic markers used in forensics to calculate likelihood ratios (LRs) of potential parent/child and sibling relationships between each offender in the a simulated database and a crime scene DNA profile. They did this to assess the chance that a crime scene sample belongs to the close relative of a cataloged offender. The higher the LR is, the more likely that the crime scene DNA sample belongs to a close relative in the offender database. “The simulations illustrate that kinship searching used in conjunction with follow-up analysis, preferably using Y-chromosomal–markers, can expand the reach of the DNA databases not by handfuls, but by gross amounts,” said Brenner.
Researchers note two important caveats, however; that although unusual, a high LR may be a result of a coincidence for unrelated individuals and that a DNA match does not necessarily prove guilt.
“Kinship analysis, as with any investigative technique, may lead to the investigation of the innocent,” says Lazer. “But it would also provide decisive leads in cases that would otherwise go unsolved. Conceivably it could provide a large number of cold hits.” A cold hit refers to the chance of finding a match between a crime scene sample and someone in the offender database.
Kinship analysis would also reflect the current composition of the database. “The database is disproportionately African American, Hispanic, poor, and those who would be indirectly included through kinship analysis would reflect those disparities,” says Lazer.
Opposition to DNA databases has been in the form of protection against civil liberties, specifically privacy. However, courts thus far have ruled that public safety concerns outweigh individual privacy rights regarding offender DNA databases. Current laws, both state and federal, do not address kinship searches, probably because these strategies were not considered when the original laws were drafted.
“While it is exciting and important to acknowledge the potential value that genetic kinship analysis could have in forensic investigations, it is also necessary to take heed of the legal and policy implications,” says Bieber.
These and related issues will be discussed at a DNA Fingerprinting and Civil Liberties National Symposium being held in Boston, May 11-13, sponsored by the American Society of Law, Medicine and Ethics.
Posted by David Lazer at May 11, 2006 10:04 PM