When a crime scene sample has evidence of more than one individual crime labs attempt to assign the results to particular individuals. When there are DNA alleles from only two people the process is usually straightforward — there is a “major contributor” and “minor contributor” identified. The problem becomes more complicated if there is evidence of DNA from three or more persons in the mixture. Until recently, many labs reported results as “inconclusive” because separating the mixture of DNA was too subjective. Then several software companies developed algorithms that could assign the DNA profiles to particular individuals by using statistical probabilities. This method of “probablistic genotyping” has been controversial among scientists in the field and forced courts to confront much new evidence.
The technology has been in use for less than 10 years, although the underlying statistical models have been around for decades in other fields. For prosecutors, the technology is an exciting opportunity to get more information to work with.
Many defense attorneys say that using the software without allowing the defense to understand how it works is unconstitutional. Every defendant has the constitutional right to fully confront the evidence against them — and that can’t be done against a “black box.” They want the source code to be made publicly available, at least.
The problem with that is these programs are owned by private companies who have a financial interest in keeping their source code and algorithms secret.
The companies providing the analysis say that turning over the source code is unnecessary because the outcomes of the analyses have been scientifically validated. However, that validation has been almost entirely performed in-house by the companies themselves.
Another problem is that defendants can’t properly confront probabilistic genotyping evidence because there are too few experts qualified to interpret it. Even if the source code were made public (and some has), few defendants can find an expert witness able to challenge, or even understand it.
Independent review finds probabilistic genotyping problematic
The Presidential Council on Science and Technology produced a report in 2016 on whether probabilistic genotyping should be admissible as evidence. It found that the technology was better than a subjective interpretation of the data.
It also found that different probabilistic genotyping programs “can yield different results for the same mixture profile.” The report said that careful scrutiny is necessary before we can determine that the method is itself scientifically valid and that the programs correctly implement the method.
Courts’ reactions to the reliability of the evidence have been mixed. At least one has ruled that the source code must be made available to the defense.
Is it fair to use probabilistic genotyping against defendants who lack the ability to challenge it seriously?
The technology has been in use for less than 10 years, although the underlying statistical models have been around for decades in other fields. For prosecutors, the technology is an exciting opportunity to get more information to work with.
Many defense attorneys say that using it without allowing the defense to understand how it works is unconstitutional. Every defendant has the constitutional right to fully confront the evidence against them — and that can’t be done against a “black box.” They want the source code to be made publicly available, at least.
The problem with that is these programs are owned by private companies who have a financial interest in keeping their source code and algorithms secret.
The companies providing the analysis say that turning over the source code is unnecessary because the outcomes of the analyses have been scientifically validated. However, that validation has been almost entirely performed in-house by the companies themselves.
Another problem is that defendants can’t properly confront probabilistic genotyping evidence because there are too few experts qualified to interpret it. Even if the source code were made public (and some has), few defendants can find an expert witness able to challenge, or even understand it.
Independent review finds probabilistic genotyping problematic
The Presidential Council on Science and Technology produced a report in 2016 on whether probabilistic genotyping should be admissible as evidence. It found that the technology was better than a subjective interpretation of the data.
It also found that different probabilistic genotyping programs “can yield different results for the same mixture profile.” The report said that careful scrutiny is necessary before we can determine that the method is itself scientifically valid and that the programs correctly implement the method.
Courts’ reactions to the reliability of the evidence have been mixed. At least one has ruled that the source code must be made available to the defense.
Is it fair to use probabilistic genotyping against defendants who lack the ability to challenge it seriously?
