Today, EDRM released a final version of guidelines for navigating technology assisted review (TAR). As one might expect from one of the e-discovery market's leading standards organizations, the document isn't exactly a beach read.

Three teams consisting of 50 volunteer e-discovery specialists, judges and practitioners helped draft the guidelines, which span 50 pages and touch on everything from quality control to the factors that lawyers should consider when debating whether or not to use TAR.

According to a Q&A shared with Legaltech News from John Rabiej, deputy director of the Bolch Judicial Institute at Duke Law, the technology can save valuable time and money but has yet to be adopted on a wide-scale by the legal community. EDRM was acquired by what is now the Bolch Judicial Institute at Duke Law in 2016.

“There simply is no good reason why millions of dollars need to be continually wasted using traditional discovery means. Changes in legal tradition come slowly, but we can no longer procrastinate when discovery costs continue to explode, while effective technology lies idle. The bench and bar need a push, and we hope to provide it,” Rabiej said.

The guidelines assert that document review is often the single largest expense associated with litigation-related discovery, accounting for 60 to 70 percent of the total cost. TAR can help expedite the document review process without doubling down on human resources.

Mike Quartararo, founder and managing director of eDPM Advisory Services and leader of one of the three teams who worked on the EDRM guidelines, wanted to make sure that message didn't get lost in scientific jargon.

Quartararo's team worked primarily on the first third of the document, which provides an overview of the TAR process. He wanted to make it as accessible as possible.

“One of our primary goals is to demystify the process. We took pains to write in plain English,” Quartararo said.

He doesn't think that TAR is even close to being used to its fullest capacity within the legal industry. Quartararo said the the goal behind the guidelines was to “educate and enlighten” and promote the use of the technology.

“One of the things we've always struggled with is why? Why wouldn't people use it?” Quartararo said.

The EDRM guidelines distinguish TAR algorithms into two categories: feature extraction algorithms and supervised machine learning algorithms.

Feature extraction algorithms are seldom tweaked by humans and analyze documents within a set and extract meaningful elements that are then organized according to value. Supervised machine learning algorithms allow a human reviewer to teach the software what is considered relevant.

The latter usually involves a technician identifying samples of relevant documents and feeding them to the computer. It's still faster than having a human attempt to manually undertake the entire review process — and possibly more accurate too. According the EDRM guidelines, studies have shown a discrepancy rate as a high as 50 percent among people who use linear review to identify relevant documents.

That doesn't mean that the computer is foolproof, though. EDRM's guidelines provide tips for quality control, such as having human reviewers preemptively identify a small pool of relevant documents that can be compared to the results the computer generates. They also lay out a suggested TAR workflow and questions to consider when evaluating a potential service provider.

EDRM will continue its TAR push within the months to come. The organization is working on developing a protocol that that dictates when and how TAR should be used. An invitation-only conference to be held in Washington D.C. this June will also examine the 2015 amendments to federal discovery rules.