Troutman Sanders LLP: 'Training' The System: The Benefits, Risks And Practical Applications Of Predictive Coding

By Troutman Sanders LLP

If you are ever responsible for responding to discovery in a lawsuit, by now you may have heard of something known as "predictive coding." Touted by many as a "game changer" in the realm of electronic discovery, predictive coding combines people and technology in a way that proponents say allows for faster, more cost-effective, and maybe even more accurate and reliable document review. However, predictive coding is not without detractors.

Predictive coding technology works by allowing attorneys to "train" their computers to find relevant documents. The first step in the process is for a reviewing attorney to review and code a few seed sets of randomly selected documents. "Coding" simply means to digitally mark a document as responsive or unresponsive to discovery requests or relevant to a certain issue in a case. The computer then takes these coded seed sets and "learns" what documents the attorney is looking for. Perhaps obviously, this process is iterative - the more computer-predicted results are reviewed and corrected by the attorney, the more accurate the predictive coding by the computer will be.

Some doubt the legal defensibility of predictive coding. Generally, the issue is with whether the use of this technology satisfies an attorney's obligation to conduct a reasonable search for responsive documents under applicable discovery rules. Numerous judges in the United States have spoken positively about predictive coding. See, for example, Judge Pecks' oft-cited decision in Da Silva Moore v. Publicis Group, et al. [enhanced version available to lexis.com subscribers]. However, until recently no judge had ever ordered its use.

That changed when Judge James Chamblin of the Circuit Court of Loudon County, Virginia became the first judge ever to order the use of predictive coding in April 2012 in the case of Global Aerospace, Inc. v. Landow Aviation, L.P. [enhanced version]. In this case, the defendants wished to use predictive coding to locate and retrieve responsive documents from a collection of 250 gigabytes of reviewable ESI. That's the equivalent of about 225 million pages of plain text. The plaintiffs objected, arguing that predictive coding technology is simply not as effective as purely human review and would only result in 75% of responsive documents being produced. Judge Chamblin sided with the defendants, ordering that the defendants were allowed to use predictive coding for the purpose or processing and producing ESI.

This decision is certainly a victory for proponents of predictive coding in that it further validates the technology as a viable review tool. But what Judge Chamblin actually did was simply extend the application of a long-standing rule, which is that litigants have the right to choose the review method that they use to review documents and ESI. It is unlikely that many people would argue that litigants need to seek the court's permission before choosing whether to use keyword searches or concept searches or to opt for manual linear review. Judge Chamblin's opinion may give added defensibility to a choice to use predictive coding, but that choice won't be deemed per se reasonable in any case. The litigants must still act reasonably in applying predictive coding technology and, most important, be able to defend their discovery conduct.

Finally, we should keep in mind that predictive coding has many uses other than locating and retrieving documents responsive to discovery requests, and these other uses generally are not subject to a court's review or approval. For instance, it can be used at the outset of a case for early case assessment. It can also be used to prioritize documents for review and to identify and isolate potentially privileged documents. It can be used as a quality assurance method to check the coding decisions of attorney reviewers. Finally, it can be used to review and analyze productions received from opposing sides.

As the amount of data subject to review in complex litigation continues to explode, new technology like predictive coding and reasonable judicial opinions like Judge Chamblin's offer hope that discovery can be effectively managed so that our civil litigation system isn't completely crippled by the costs of document review. Cases should be resolved on their merits, not because they are just too expensive to litigate. Hopefully, more wide-spread acceptance of predictive coding can help protect that principle into the future.
For more information, please contact Erin Graham, Alison Grounds or John Hutchins.

­­­­­­­­­­­­­­­­­­­­About Troutman Sanders

Troutman Sanders is an international law firm with offices in North America, Europe and Asia. Founded in 1897, the firm's heritage of extensive experience, exceptional responsiveness and an unwavering commitment to service has garnered strong, long-standing relationships with clients across the globe. These clients range from multinational corporations to individual entrepreneurs, federal and state agencies to foreign governments, and non-profit organizations to businesses representing virtually every sector and industry.

Troutman Sanders lawyers provide counsel and advice in practically every aspect of civil and commercial law related to the firm's core practice areas: Corporate, Finance, Litigation, Public Law and Real Estate. With more than 50 practice groups focused on specific aspects of these areas, the firm is defined by its considerable knowledge base and proactive approach to addressing legal and business challenges.

For more information about LexisNexis products and solutions, connect with us through our corporate site.