Last month, in Rio Tinto PLC v. Vale S.A., No. 14 Civ. 3042 (RMB)(AJP) (S.D.N.Y. Mar. 3, 2015), Magistrate Judge Peck issued another order regarding the use of predictive coding (a/k/a technology assisted review, or TAR). Since then, more words have been written about his order than are in it. Why the fuss? What’s the significance of the decision? For starters, Judge Peck’s titling of the order “Da Silva Moore Revisited” caused many in the e-discovery community to take notice.
But Rio Tinto is not groundbreaking. Judge Peck approved the parties’ agreed-upon protocols for using predictive coding, which a number of other courts have done since Da Silva Moore—no big deal. More notable is Judge Peck’s statement of several generalized principles about predictive coding. Most experienced e-discovery practitioners already followed these principles, but it’s nevertheless helpful to have an influential federal magistrate judge articulate them. So here’s my recap of the order’s highlights.
“It is now black letter law that where the producing party wants to utilize TAR for document review, courts will permit it.”
In the three years following Judge Peck’s decision in Da Silva Moore—the first published federal court order approving the use of predictive coding—case law has developed enough for him to declare that the acceptability of predictive coding is “black letter law.” Pages 2–4 of his order collect the case law to date.
“While I generally believe in cooperation, requesting parties can insure that training and review was done appropriately by other means.”
According to Judge Peck, one still-open issue around predictive coding is “how transparent and cooperative the parties need to be with respect to the seed or training set(s).” To be sure, that issue still generates much discussion. But I do not believe that a court has the authority to require the disclosure of seed sets or other documents that do not fall within the scope of discovery under Rule 26, or to force a certain level of cooperation, so long as the parties otherwise comply with their discovery obligations. That said, a reasonable degree of cooperation can ensure a more efficient and cost-effective discovery process, and parties may want to consider volunteering certain information about their use of predictive coding to move the process along. (Take a look at my last blog post, in which I discuss sharing non-responsive documents in the context of search term optimization and negotiation.)
However, as Judge Peck notes, there are other ways to validate a predictive coding process (or its results) besides disclosure of seed or training set documents. We advocate statistically valid measurements of recall and precision. But regardless of how a litigant chooses to ensure the defensibility of its TAR process, keep in mind the final highlight of Judge Peck’s order. . .
“[I]t is inappropriate to hold TAR to a higher standard than keywords or manual review.”
Bravo, Judge Peck. At DiscoverReady, several of our clients have expressed reluctance to use predictive coding for fear that they would spend more in motion practice—or protracted negotiations with opposing counsel—than they would save by using the tool. Holding predictive coding to a higher standard of defensibility or transparency than other discovery processes will discourage its use. So long as the process is reasonable—the operative standard for discovery disclosures—it should not matter what tools or techniques the party used. And no party should be required to prove up the defensibility of its processes unless and until the other side has shown some deficiency in the production.
Hopefully other courts will adopt Judge Peck’s perspective on the use of TAR and other advanced analytics that improve the efficiency and accuracy of document discovery. As the volumes of information increase, parties soon will have no choice but to deploy these tools—and they should not be forced to meet a heightened standard of defensibility when they do so.