Predictive Coding: Green Light or Yellow Caution?

It’s not often that a discovery order is so eagerly anticipated that the judge has to calm down an overheated blogosphere. This week’s hotly discussed e-discovery ruling from U.S Magistrate Andrew Peck regarding the use of predictive coding in litigation includes an unusually direct footnote for legal bloggers. “To correct the many blogs about this case, initiated by a press release from plaintiffs’ vendor– the Court did not order the parties to use predictive coding,” he wrote. “The parties had agreed to defendants’ use of it, but had disputes over the scope and implementation, which the Court ruled on, thus accepting the use of computer-assisted review in this lawsuit.”
This is arguably the first court order on the use of machine learning, or predictive coding search technology in litigation. The case, Da Silva Moore v. Publicis Groupe, No. 11 Civ. 1279 (S.D.N.Y. Feb. 8, 2012), is a gender discrimination suit involving approximately three million electronic documents. The parties had been battling over how to best review these documents and narrow the number of documents entered into evidence.
In this case, machine learning is making it possible for computers to assist in the relevancy review process by recognizing responsive documents in eDiscovery. With the application of advanced machine learning algorithms, machines have been found to be as effective as human reviewers. In this case, the plaintiffs had been reluctant to allow the liberal use of predictive  coding in review, but Judge Peck ultimately stepped in and set the parameters of how computers should be deployed. He noted in court that unless he did so, “it will be five years before discovery is concluded, because each of you doesn’t like what the other is doing.”
Peck’s order provided that human reviewers should look at a sample set of documents and code them as relevant or irrelevant, allowing for privileged documents to be set aside. Then, using responsive documents to help train the computers to find other responsive documents, the court will allow for seven rounds of computer review and human quality-checking of random samples to narrow the collection.
Despite Peck’s endorsement of machine learning technology in litigation, nothing is truly settled, and the battle over technology-assisted review may only get more contentious. The plaintiffs actually went so far as to incorporate a paragraph into the stipulation and order stating that they “object to this ESI Protocol in its entirety,” and “reserve the right to object to its use in the case.”
However, tech savvy lawyers can still hope for a stronger endorsement of machine learning technology. The case Kleen Products LLC v. Packaging Corporation of America, et al., is an antitrust matter in the United States District Court for the Northern District of Illinois where an e-discovery order is also anticipated. In this case, the plaintiffs seek a court order requiring defendants, among other things, to use predictive coding technology to respond to plaintiffs’ document requests.
Like anything in the law, it will likely take years before the there is a full endorsement of technology-assisted review, and the details of how it will be deployed will always be subject to vigorous debate. However, Da Silva Moore is clearly a move in the right direction.