July 12, 2012 Top Story

Predictive Coding Portends Change in Discovery Practice

The courts' acceptance of computer-aided search could reduce review costs

Swati S. Desai

The use of predictive coding in large-scale complex litigation is slowly becoming more mainstream. Two recent court decisions on this next step in e-discovery practice address the issue of whether predictive coding can be used even when parties disagree about the keywords used, custodians searched, and the acceptable degree of accuracy. DaSilva Moore v. Publicus Groupe SA [PDF]; Global Aerospace Inc. v. Landow Aviation, L.P. [PDF].

To many, the concept of predictive coding remains elusive. “The biggest issue that bothers me with predictive coding is that people don’t know what it is. We don’t have a definition,” observes Danielle M. Panetta, Boston, Newsletter Editor for the ABA Section of Litigation’s Technology for the Litigator Committee. “Attorneys up until three years ago were still debating, ‘What is a document?’ Right now there are different definitions for a document in different jurisdictions. Given that, how can you even begin to describe predictive coding, with its training requirements and algorithms?”

Reigning Forms of Document Review

To understand predictive coding, one needs to understand what it is not. It is not human review, which involves taking a stack of documents and handing them to an associate, paralegal, or contract attorney for manual review. This type of review takes the most amount of time and can become very expensive very quickly. It is best used for small-scale document reviews. A purely “human search is rarely done,” notes Aaron H. Gould, Newark, cochair of the Electronic Discovery Subcommittee of the Section of Litigation’s Pretrial Practice & Discovery Committee. Technological aids, such as date filters, often accompany even a manual review.

It is also not keyword-search review, which involves unilaterally selected, agreed-on, or court-ordered keywords to search designated custodians’ files. Keyword searches aim to lessen the time and expense associated with manual review by narrowing the scope of the search. “The idea is that documents that are relevant to the case would contain those keywords,” explains Gould.

Predictive coding is different. Human reviewers review an initial set of documents. A software program records and analyzes their responsiveness determinations, then designs algorithms that, in lieu of manual reviewers, make subsequent responsiveness decisions. Reviewers test and refine the algorithms to verify the program’s accuracy at retrieving responsive documents.

Detractors question whether predictive coding satisfies a party’s obligation to perform a reasonable search for responsive documents. Recent decisions favoring predictive coding observe, however, that it can be more effective at identifying responsive documents than manual or keyword review.

In DaSilva Moore, Magistrate Judge Andrew J. Peck explained his support for predictive coding while ruling on a party’s motion [PDF] to remove him from the case. He stated that “key words, certainly unless they are well done and tested, are not overly useful. Key words along with predictive coding and other methodology, can be very instructive.” Judge Peck also indicated that for predictive coding to be allowed during discovery, the seed set, including nonresponsive documents, should be provided to enable counsel and the court to ascertain whether the search parameters and algorithm are reasonable and whether keyword or human search would work better. It was this measure of the reliability of the predictive-coding process that encouraged Judge Andrew L. Carter to uphold Judge Peck’s ruling.

Panetta emphasizes that predictive coding does not involve just feeding documents into a computer and letting it make discovery decisions that counsel should be making. “You are spending a lot of attorney time on analysis before employing” the software program, cautions Panetta. “You have to alter what you are doing based on substantive legal analysis.”

Potential Benefits of Predictive Coding

“Predictive search will reduce the costs of discovery by one or two orders of magnitude,” predicts Richard S. Stockton, Chicago, cochair of the Section’s Technology for the Litigator Committee. “Predictive search would revolutionize e-discovery if fully implemented and accepted.”

Stockton often defends clients in patent-infringement actions where discovery costs can run very high. “It would take millions and millions of dollars to complete discovery. Instead, it might make better business sense to . . . settle and just be done with it,” says Stockton. “With predictive coding, discovery costs decrease from several millions of dollars to a few hundred thousand dollars to tens of thousands of dollars. Corporate defendants are more likely to fight because, once you get through discovery, you can address the case on its merits.”

Pitfalls of Predictive Coding

Recent years have seen dramatic growth in the number of electronic-discovery vendors, as corporations’ data sets have surged from megabytes to terabytes. Many vendors provide contract attorneys to perform manual-search or keyword-search review. Widespread use of predictive coding could shrink this segment of the industry. “Review vendors might have some difficulty surviving if predictive coding goes mainstream. They might have to change their” business plan, warns Stockton. “The days of a vendor hiring 50 attorneys to do a short-term document review are numbered.”

The Future of Predictive Coding

The field of predictive coding is still very new, and only three court cases have specifically approved the use of predictive coding where there was opposition from opposing counsel. As more cases wind through the court system, the legal community will have a better idea of whether predictive coding obtains widespread acceptance as a discovery tool and whether the costs for human and keyword search can be shifted to the other party if they do not want to use predictive coding.

Swati S. Desai is an associate editor for Litigation News.

Keywords: discovery, predictive coding, electronically stored information, ESI

Related Resources

  • Da Silva Moore v. Publicis Groupe et al., No. 11 Civ. 1279, 2012 U.S. Dist. LEXIS 23350 (S.D.N.Y. Feb. 24, 2012).
  • Global Aerospace Inc. v. Landow Aviation, L.P., (Va. Cir. Loudon Co. April 23, 2012 ) (Consolidated Case No. CL 61040).

Copyright © 2017, American Bar Association. All rights reserved. This information or any portion thereof may not be copied or disseminated in any form or by any means or downloaded or stored in an electronic database or retrieval system without the express written consent of the American Bar Association. The views expressed in this article are those of the author(s) and do not necessarily reflect the positions or policies of the American Bar Association, the Section of Litigation, this committee, or the employer(s) of the author(s).