Forensic science, like many other fields, is beginning to see the potential impact of computer processing on its standard methodologies. But questions loom about the viability and strength of these methods in making finite determinations, especially when the freedom of criminal defendants is on the line.

Speakers at Friday's 2018 Georgia State Law Review Symposium, titled “From the Crime Scene to the Courtroom: The Future of Forensic Science Reform,” sought to address some of the ways in which technological advances have helped debunk old forensic science and created new challenges for criminal lawyers. One of the symposium's panels, “Technological Advancements in Forensic Science,” considered the potential benefits and ongoing concerns around applying advanced computing methods to DNA forensics used in criminal evidence.

Panel moderator and GSU College of Law associate dean Jessica Gabel Cino framed the discussion by poking some holes in the perceived accuracy of DNA testing, a scientific method that has cleared the names of a number of wrongfully convicted individuals. “DNA is seen as that gold standard that tends to equalize things,” she said. “It's not necessarily that straightforward. It can be very complex when you're talking about, for example, mixture cases,” where multiple people's genetic material is represented in the same DNA sample.

New forensic technology, she noted, attempts to use algorithms to bring clarity to those mixture cases. Companies touting software solutions to these problems—such as TrueAllele Casework, which has software slated to be adopted by the Georgia Bureau of Investigation—often do not disclose their algorithms, making it difficult for attorneys to interrogate their methods. 

Clint Hughes, senior staff attorney at the Legal Aid Society of New York City, worked on a few cases where software from companies like TrueAllele and STRmix have produced some discordant results from one another. While he found that the software does seem to have a good track record in deducing variation in DNA, there can certainly be reason to interrogate their methodology and results.

“These kinds of software are more acceptable to the experts that we used because they take the peak heights into account, and they take other kinds of modeling into account, but that does not mean that they are necessarily offering accurate results,” Hughes said.

Forensic Bioinformatic Services systems engineer Nathan Adams noted that, while previous forensic science attempted to simply identify the presence of a particular person in a DNA sample, current technology is looking instead to identify where and how much of a person is represented in those samples. “We're moving in a direction of more discriminatory power with these more complex programs,” he said.

While there is some published research and source code available, these methods are typically proprietary, making it tough for attorneys to tell exactly how these methodologies are being applied. “We don't necessarily know what's going on behind the scenes,” Adams said.

ACLU legal fellow Vera Eidelman urged audience members to consider what incentive structures make it difficult for attorneys to gain access to the information they might want or need about these algorithmic methods. “These are proprietary algorithms that are being put out by private companies that are attempting to make a profit, and are now claiming private property interests. They're claiming that the underlying algorithms are trade secrets,” she said.

“There are just more barriers to finding out all of the important things that have been discovered for other types of forensic science that we've been hearing about throughout the day,” Eidelman said, nodding to the previous panels' discussions of now-discredited methods of attribution.

South Dakota State University math and statistics professor Cedric Neumann acknowledged that the science used by companies like TrueAllele and STRmix has been published and peer reviewed. However, attorneys are faced with attempting to deduce whether the software applies this science appropriately and whether the software has applied to the particular sample effectively.

Chuck Boring, Cobb County deputy chief assistant district attorney, said jurors are fairly wary of being duped by supposedly complex technology. “You have to be upfront about the limitations,” he said.

However, acknowledging the limitations of the science to jurors may mean that prosecutors can't hinge their case on this method alone. “It may be more useful to use it as corroboration,” Boring said.

Although these new potential uses of computing and algorithmic learning have great potential to shift forensic science, Adams said scientists still have quite a way to go before these methods can be really made indisputable. “We've started to talk about it,” Adams said, “But we're nowhere near what I would consider to be a mature state of having fully incorporated computing as a relevant discipline that we need to cross-train on.”