Manual review of ESI has become nearly impossible with the increasing volume of electronic information avalable, leading many companies to ask “what is predictive coding and should we use it?” Based on a recent article by U.S. Magistrate Court Judge, Andrew Peck, which was published in Law Technology News, the answer is yes – there are cases thatshould take advantage of predictive coding in e-discovery.
Predictive coding is also known as technology-assisted coding. In predictive coding, documents are reviewed by one or a team of senior partners and a “seed set” of documents is coded using a specialized computer program. As documents are coded, the system begins to understand the meaning of each code. In some cases, the reviewers are asked for feedback from the system to insure validity. Once the reviewer’s coding and the computer’s predictions coincide, it is determined that the system can confidently predict coding for the remaining documents to be reviewed.
Other judges seem to agree with Judge Peck in supporting the use of technology-assisted coding. In Disability Rights Council of Greater Washington v. Washington Metro Transit Authority 242 F.R.D. 139 (D.D.C. 2007), Magistrate Judge John Facciola pointed out that concept searching is more likely to produce comprehensive results and is more efficient than keyword searches. However, the question remains as to whether litigants can demonstrate that the results are defensible to the courts.
Judge Peck indicates that if the litigants could explain what was done and demonstrate that predictive coding provided reliable results with reasonably high recall and precision, he would approve the use of the system. In addition, Magistrate Judge Paul Grimm of the District of Columbia stated in Victor Stanley, Inc v. Creative Pipe, Inc. 250 F.R.D. 251 (D. Md. 2008) that litigants should expect to support the method they use for search and retrieval with affidavits or testimony from qualified individuals. Therefore, when using predictive coding it is imperative to document how the coding was done, to validate the results, and to have a competent team member available to explain the methodology.
No Judicial Endorsement of Keyword Searches
For many who are asking “what is predictive coding?” the belief is that judges have sanctioned the use of keywords, and new technology like predictive coding is going to open them up for challenges. Some point to comments made by Judge Facciola in a later decision, United States v. O’Keefe, 37 F. Supp. 2d 14 (D.D.C. 2008), where he discussed the importance of compliance with Rule 702 of the Federal Rules of Evidence. However, as Judge Peck points out, that discussion was in the context of a challenge to the sufficiency of keywords. According to Judge Peck, the bench has actually criticized keyword searches quite often, including opinions by Judges Grimm and Facciola.
Despite the fact that many companies believe that keyword searches with manual review are the best method for seeking electronic documents, two recent studies have shown otherwise. Researchers for the Electronic Discovery Institute, as well as Maura Grossman, a litigation attorney, and Gordon Cormack, a university professor, conducted studies comparing human review to predictive coding and found that technology-assisted review provided equal or better results than manual review with less effort. In addition, because keyword searches issue many false positives, litigants must use expensive manual review to find documents that are relevant. For this reason, firms must not only ask themselves “what is predictive coding” but “when should I begin using predictive coding?”