Move Over Humans, 21st Century Document Review Has Arrived
Beyond highlighting the evolving standards and complexities of establishing a comprehensive e-discovery plan at the early stages of a case, recent decisions in Da Silva Moore v. Publicis Groupe (S.D.N.Y.), which approved a defendant’s use of predictive coding, could potentially have a Zubulake-type effect of establishing new technology-driven norms for large-scale document reviews in the future.
The Da Silva case was brought as a putative $100 million class and collective action where the plaintiffs alleged they were subject to systemic gender discrimination in violation of Title VII of the Civil Rights Act of 1964. Because of the substantial volume of electronically stored information (ESI) proposed to be collected by Publicis and its U.S.- based public relations subsidiary, MSL Group, both parties had agreed to the concept of using computer-assisted review. Sources of ESI were identified, multiple phases of discovery were established, and a detailed discovery protocol was ordered by the Court. Plaintiffs later objected to the particulars of the predictive coding protocol and even its use, arguing that it was unreliable and contrary to F.R.E. 702 and the Supreme Court’s Daubert decision.
Predictive coding software uses sophisticated algorithms to enable a computer to determine relevance based on how well the documents match concepts and terms in "seed documents" selected by human reviewers. Da Silva is one of the first cases in the country in which a court has weighed in on the use of predictive coding. Over the plaintiffs’ objections to the particulars of the protocol for the defendants’ use of predictive coding, Magistrate Judge Andrew Peck ordered a particular protocol to be followed initially while preserving the right of the plaintiffs to raise further issues based on the results. The plaintiffs appealed to District Judge Andrew Carter, arguing, among other things, that Magistrate Judge Peck’s public articles and speeches evidence a bias in favor of predictive coding and the defense bar. Judge Carter rejected the appeal as premature while echoing Magistrate Judge Peck’s inclination to accept predictive coding as a reasonable method to cull through millions of pages of potentially relevant ESI.
Judge Carter took the same "wait and see" approach as Magistrate Judge Peck regarding the reliability of the process, explaining (1) the majority of documentary evidence has to be produced by the defendants, (2) even if all parties were willing to entertain the notion of manually reviewing the documents, such review is prone to human error and marred with inconsistencies from the determination by various attorneys of whether a document is responsive, and (3) if the method provided in the protocol does not work, or if the sample size is indeed too small to properly apply the technology, the court will not preclude plaintiffs from receiving additional relevant information. Nothing precluded the plaintiffs from revisiting the issue of the software’s efficacy or reliability at a later phase in discovery, and thus, the plaintiffs’ arguments concerning the reliability of the method were premature. Other courts seem to be following this trend – see Global Aerospace, Inc. v. Landow Aviation (state court in Virginia allowing defendants to use predictive coding for processing and producing ESI – decided April 23, 2012).
While further lessons regarding the use and reliability of predictive coding will likely be developed in Da Silva as that case progresses, here are a few immediate takeaways:
- Computer-assisted review may not be appropriate in all cases, but given the need for cost effectiveness and proportionality under F.R.C.P. 26(b)(2)(C) and (5), the parties should at least discuss this type of review as an option in large data-volume cases.
- Vendors and/or counsel should be able to explain complicated e-discovery concepts in ways that make it easily understandable to judges who may not be tech-savvy, and should be able to work with opposing counsel to help determine the best ways to implement these types of technologies.
- To control discovery costs, consider staging discovery in phases by starting with the most likely to be relevant sources (including custodians), without prejudice to the requesting party seeking more after conclusion of that first stage review.