Breaking News: Judge Peck Declines to Compel TAR, Provides More Guidance on TAR Models

Adam Kuhn
August 3, 2016

When Judge Peck speaks, the eDiscovery community listens. Yesterday, he issued a new order in Hyles v. New York City addressing an issue that has surfaced before in case law: Can a requesting party compel the producing party to use TAR? While the short answer is no, the longer answer provides some food for thought on the future of legal technology use and the potential for mandatory TAR. The Keyword Search Incumbency Hyles deals with claims of employment discrimination and features a contentious discovery history. Initially, the parties could not agree on custodians and date ranges, and the court issued rulings that “largely accepted the City’s [narrower] scope parameters.” After a discussion about keywords, the requesting party proposed that the City use TAR for review. The City refused, preferring keywords, and the matter was referred to Judge Peck to decide whether the City could be compelled to use TAR. The court endorsed TAR generally as “cheaper, more efficient and superior to keyword searching” and “the best and most efficient search tool” for most cases. That Judge Peck approves of TAR is no secret; he has authored two of the most influential opinions solidifying the defensibility of Predictive Coding. His jurisprudence, it is fair to say, has helped pave the way for broader adoption of legal technology. But, as he acknowledges in this latest opinion, we are not yet at a point “when TAR is so widely used that it might be unreasonable for a party to decline to use TAR.” And so the court rightly denies the request to compel the use of TAR—finding in favor of the producing party’s discretion to adopt a reasonable method of their choosing.

“The Court would have liked the City to use TAR in this case. But the Court cannot, and will not, force the City to do so.”

Judge Peck Panel (Judge Peck [second from left] presenting on a Legaltech West Coast panel)

Sedona Principle 6 Carries

Judge Peck acknowledges and discusses this tension between competing ideals in the law. On the one hand, courts have lamented the lack of cooperation between parties. But on the other hand (and referencing the oft-cited Dynamo case) courts are not in the business of “dictating to a party the manner in which it should review documents.” Facilitating efficient discovery without co-opting discovery has proven to be a challenging line to walk. That said, courts have almost unanimously sided, as encouraged by Sedona Principle 6, in favor of the producing party’s right to use their preferred technology. Most of the cases on this topic deal with a party that wants to use TAR but had already negotiated a protocol otherwise. But there are other, more unusual instances where the issue of compelled TAR usage has been raised. For instance, in Good v. American Water Works, the requesting party attempted to make their agreement on a 502(d) order contingent on the producing party’s use of TAR for expedited privilege review where they wanted to conduct a lengthy manual review. While the argument was ultimately rejected, the court did genuinely discuss it and even left it on the table as an option in case of deficient production.

Does Continuous Learning Obviate these Issues?

One reason that judges, including Judge Peck, are reluctant to compel the use of TAR is that the traditional TAR 1.0 model relies on across-the-aisle collaboration. In Hyles, the producing party resisted adopting TAR because there had been little agreement and “based on their history of scope negotiations, would not be able to collaborate to develop the seed set for a TAR process.” In his latest post on the new Dynamo Holdings opinion, Hal Marcus discussed the limits of cooperation through the lens of Dynamo II and raised a salient point about choosing the right TAR model for the job. There, Hal wrote that many of the issues could have been avoided by using a continuous machine learning system. In Hyles, Judge Peck shares a similar sentiment:

“To be clear, the Court believes that for most cases today, TAR is the best and most efficient search tool. That is particularly so, according to research studies (cited in Rio Tinto), where the TAR methodology uses continuous active learning (“CAL”), which eliminates issues about the seed set and stabilizing the TAR tool. . . There may come a time when TAR is so widely used that it might be unreasonable for a party to decline to use TAR.” 

As litigants and courts become more comfortable with TAR generally, we will undoubtedly see this line of analysis into the specific TAR methodology applied more frequently. Given the recent movement at the state and federal level towards requiring increased technological competency for attorneys, that time may come sooner than later.

Solution Brief: Predictive Coding Differentiated