Judges approve predictive coding for e-discovery

Text Size

Share this page on any of these social networking sites:
Share this page on any of these social networking sites: LinkedIn

 

Advertise with Trial

Stand out from the crowd! Advertise in Trial  to reach a national audience of major decision-makers who are looking for products and services to improve their legal practices.

Learn more »

Top Story

June 7, 2012

Judges approve predictive coding for e-discovery 

Steven M. Sellers

Predictive coding—a computer program that uses algorithms to search and produce multiple documents based on their relevance—may be used for voluminous e-discovery in New York and Virginia cases. Some lawyers have expressed concern about using predictive coding in the absence of a judicial decision approving it.

A federal magistrate in New York and a state judge in Virginia allowed the use of predictive coding to search electronically stored information in two recent cases, signaling how other courts may deal with discovery of voluminous electronic data. Predictive coding is a computer program that uses algorithms to search and produce multiple documents based on their relevance. Similar in concept to an email spam filter, the program sifts through documents to select those most relevant. The program learns to recognize relevant documents in one of two ways: through an exemplar of documents chosen by a human or one chosen by the computer, which is progressively graded by a human for relevance as the program conducts the search.

Proponents contend that predictive coding is the next generation of software to replace keyword searches, which sometimes produce irrelevant or incomplete results. Corporate defendants often favor predictive coding as a cost-effective way to cull from mountains of electronic data for discovery requests, but some lawyers have expressed concern about using the protocol in the absence of a judicial decision approving it.

In Monique Da Silva Moore v. Publicis Groupe, a putative gender discrimination class action pending in the Southern District of New York, Magistrate Judge Andrew Peck approved the use of predictive coding for discovery, noting that “this opinion appears to be the first in which a court has approved of the use of computer-assisted review.” While acknowledging that the method may not be appropriate in all cases, Peck said that “computer-assisted review is an available tool and should be seriously considered for use in large-data-volume cases where it may save the producing party (or both parties) significant amounts of legal fees in document review.”(Monique DaSilva Moore v. Publicis Groupe, No. 1:11-cv-01279-ALC-AJP (S.D.N.Y. Feb. 24, 2012).)

Peck’s ruling stated that the plaintiffs agreed to the use of predictive coding in the case, but the plaintiffs argued that his ruling mischaracterized the scope of their acquiescence and filed an objection with U.S. District Judge Andrew Carter. Carter adopted Peck’s order, noting that “there is simply no review tool that guarantees perfection.” He added that Peck’s order allowed the parties to raise any issue that may arise in the new technology’s use, rendering the objection immaterial. “Judge Peck has ruled that if the predictive coding software is flawed or if plaintiffs are not receiving the types of documents that should be produced, the parties are allowed to reconsider their methods and raise their concerns with the magistrate judge,” Carter wrote.

Peck’s ruling—or at least its implementation—was subsequently thrown into question when he stayed it to allow resolution of the plaintiffs’ request to certify the case as a class action. That request is pending before Carter, as is the plaintiffs’ motion to recuse Peck from the case, in part because he authored an article in 2011 that praised predictive coding as a discovery tool.

Lawyer Daniel Garrie, an e-discovery neutral and special master in Seattle, said the case “evidences the importance of counsel engaging in cooperative dialog up front and early when addressing e-discovery issues,” as well as the need for lawyers to understand the technology and practical implications when large amounts of electronic data are involved. He said that when both sides have huge data sets, they are likely to be amenable to predictive coding.

But even in litigation with large data sets on both sides, predictive coding may not be a panacea if a party can establish that the method fails to produce all relevant documents. In the Virginia case, Global Aerospace, Inc. v. Landlow Aviation, at least one plaintiff objected to the use of predictive coding to search through more than 250 gigabytes of data. Circuit Court Judge James Chamblin approved the use of the discovery method but allowed any party receiving the e-discovery to raise any issue “as to the completeness or the contents of the production or the ongoing use of predictive coding.” (Global Aerospace, Inc. v. Landlow Aviation, No. CL 61040 (Va., Loudoun Co. Cir. Apr. 23, 2012).)

A request to use predictive coding is also pending in a federal case in the Northern District of Illinois. The plaintiffs in the antitrust case have asked the judge to use predictive coding based on their concerns that a previously ordered keyword search had failed to produce all relevant documents. A ruling on the motion is pending. (Kleen Prods. v. Packaging Corp. of Am., No. 1:10-cv-05711 (N.D. Ill. Sept. 9, 2010).)


The American Association for Justice
777 6th Street, NW, Ste 200 • Washington, DC  20001 • 800.424.2725 or 202.965.3500

© 2014 AAJ