News & Press: News

TAR Times Two: Dueling Rulings on Predictive Coding

Thursday, August 11, 2016   (0 Comments)
Posted by: Jason Krause
Share |

In recent weeks, two rulings on predictive coding have been published on opposite sides of the Atlantic Ocean. On the surface, the decisions come to opposite conclusions, although technology-assisted review (TAR) proponents say that both qualify as full-throated endorsements of the use of the technology in litigation.

Judicial approval of predictive coding technology is still an important question for the profession. For example, during a TAR webinar co-sponsored by ACEDS and CloudNine yesterday, when polled, almost half of the attendees believed that TAR has been ordered by US courts. In fact, the issue is still being litigated. 

The first case is Brown v BCA Trading from the UK. This opinion was actually issued back in May, but the published opinion has only become available online this month. (You can read the opinion here.) It follows an earlier legal analysis in Pyrrho Ltd v MWB Property Investments from February which laid the groundwork for using predictive coding in document review. In this case, the court ordered that computer-assisted review should be used to control the time, cost, and growing complexity of reviewing electronic evidence.

Meanwhile, in the United States, the newest published predictive coding case is Hyles v. NYC, 110-cv-03119 (SDNY Aug. 1 2016. This employment dispute involved more than five years’ worth of email and other communications collected from nine custodians. The plaintiff in the case suggested that computer-assisted review would be useful in this case, but the attorneys for New York City resisted. Despite being authored by a well-known judicial advocate for computer-assisted review in litigation, the court declined to order the technology be used in this case.

The Hyles opinion was authored by Judge Andrew Peck of the Southern District of New York, who published the first opinion authorizing predictive coding in litigation and in several subsequent cases. For example, in Rio Tinto v. Vale, et al., March 2, 2015, Judge Peck approved a predictive coding protocol, which had been suggested by the parties.

In Hyles, the plaintiff’s counsel argued that using TAR would be the most cost-effective method for reviewing several years’ worth of e-mails and other electronically stored information. However, the City’s legal team argued that the e-discovery negotiations in the case had already been so contentious that they didn’t believe the parties could cooperate sufficiently for the TAR process.

It is important to note that Judge Peck remains a supporter of TAR, writing, “It certainly is fair to say that I am a judicial advocate for the use of TAR in appropriate cases.” However, Judge Peck noted that the Sedona Principles say that, “Responding parties are best situated to evaluate the procedures, methodologies, and technologies appropriate for preserving and producing their own electronically stored information.”

Several years ago, Judge Peck wrote in Rio Tinto, “[T[he case law has developed to the point that it is now black letter law that where the producing party wants to utilize TAR for document review, courts will permit it.” However, Hyles makes it clear that the producing party can squelch any efforts to force employment of the technology as well.

Two Legal Systems Separated by a Common Language

Maura Grossman is a Research Professor at the University of Waterloo, as well as an e-discovery attorney and consultant in New York City. She says that though U.S. courts and U.K. courts differ in the level of detail and analysis applied to the question of using predictive coding in litigation, the conclusions reached are very similar.

In particular, she points out that the UK court in Pyrrho Investments developed a ten factor test to determine if predictive coding is an appropriate tool. This test would likely be familiar to many U.S. attorneys, as it relies on the same underlying principles as are found in the United States’ Federal Rules of Civil Procedure. “While the factors outlined refer to the specifics of the cases at hand, as well the specifics of U.K. law, they implicate the same principles of reasonableness and proportionality that also apply in common law jurisdictions, including the United States,” says Grossman.

In fact, judges in the UK may have an easier time imposing predictive coding protocols in a matter. Writing on his blog, UK attorney and e-discovery expert Chris Dale points out that judges there are encouraged to actively manage cases and control costs, especially since the loser pays the winner's costs in English Courts. He argues that in Hyles, the Sedona Principles were an impediment to sensible case management, which can include advanced search technology. "If Sedona Principle 6 is standing in the way, perhaps it is time for a more nuanced version, especially since last year’s FRCP amendments gave more early case management powers to the court," he writes. 

What is the Holdup?

Despite opinions such as these, adoption of predictive coding has been slow and fitful in both the U.S. and U.K. According to research by Gartner, “Predictive coding…has not gained mainstream adoption.  The estimated rate of adoption among enterprises is [only] about 10% to 15%, while the service providers may reach 50% to 60%.”

Part of the reason may be that there remain areas of contention and uncertainty even in cases where the courts have authorized the use of machine-assisted review. In Brown v BCA Trading, the court notes that the cost and complexity of the matter is likely to get worse as the case grows. “The financial factors will be different in every case, and case-to-case differences are likely more substantial than any inter-jurisdictional differences,” says Grossman. “Cost is always a consideration, but it is often challenging to accurately predict costs before one embarks on an e-discovery process.”

For example, the prevalence of responsive documents has a significant impact on the cost of a TAR review, and it is rarely known before the review process begins. However, Grossman believes these challenges will be overcome. “Looking back at history, it is not obvious precisely when automobiles supplanted horses for local transit, or airplanes supplanted trains as the most reasonable form of transcontinental transit,” says Grossman, whose research includes review processes using machine learning technology. “But those older methods were eventually replaced, and the same will be true as to keyword search and manual review. It is not a question of if, but when.”

Despite ruling that TAR would not be used in this matter, Judge Peck endorsed the use of the technology, in particular, continuous active learning or CAL as the most effective means for conducting review. “The $64,000 question is, when will CAL—or an even better method—become so universally accepted that it will be unreasonable not to use it?” asks Grossman, who has earned several patents for CAL technology.  “Judge Peck states in Hyles we are not there yet, and I cannot disagree with him on that point.”

What our customers say?

©2016 Association of Certified E-Discovery Specialists
All Rights Reserved