- “Predictive Coding” and Litigation
- May 14, 2015
- Law Firm: Pessin Katz Law P.A. - Towson Office
- Recently, PK Law’s newsletter contained a piece on the use of electronic means to effect service of process. The topic found its way into the national press which captured the decision of a judge to permit service on a defendant by means of Facebook®.
In this newsletter article, PK Law explores another intersection of law and technology, the e-discovery area of “Technology Assisted Review” (“TAR”) or “Predictive Coding”. And, so, we harken back to the year 2012 when U.S. Magistrate Andrew J. Peck issued what some commentators refer to as the seminal “TAR” ruling in Monique Da Silva Moore, et al. v. Publicis Groupe (USDC SDNY, 11 Civ. 1279 (ALC) (AJP)(2/24/12)).
Magistrate Peck was no stranger to TAR, having previously authored an article titled Search, Forward: Will manual document review and keyword searches be replaced by computer-assisted coding? (L. Tech. News, Oct. 2011). In essence, he stated in his article that there was no clear statement from the courts as to the propriety of searching Electronically Stored Information (“ESI”) and no “approved” software vendor for so doing under the Federal Rules of Civil Procedure (“FRCP”). He said those waiting for such a statement will have “a long wait”.
But what Magistrate Peck did articulate in his opinion was a view that if TAR could “secure the just, speedy, and inexpensive” (See, FRCP 1) determination of cases then it would be appropriate to employ it. In Da Silva Moore, he made clear that he “did not order the parties to use [TAR]”, but that the parties agreed the defendants could use it. The court’s role was resolving “disputes over the scope and implementation” of its use, thereby implicitly accepting TAR’s use in the case.
Da Silva Moore arose under Title VII, with allegations of gender discrimination, pregnancy discrimination, violations of the Family and Medical Leave Act, the Equal Pay Act and Fair Labor Standards Act, and similar New York state and local laws, as applicable. However, for purposes of this article the facts, and the law applicable to those facts, is not the focus of discussion. Suffice it to say that document custodians involved with the case were holding approximately three million applicable documents.
So, what is “TAR”? Magistrate Peck points out his understanding of it by citing his article’s discussion of the technique. To paraphrase, rather than have one or more junior “staff” manually review documents for relevance, a senior attorney assigned to the case, perhaps in conjunction with other staff, review a “seed set” of documents, which might number in the thousands. Using software “algorithms” developed by vendors, the computer determines their relevance to the case. The computer might then use that determination to “code” other documents for relevance. As more sample documents are reviewed, the computer “predicts” the reviewer’s coding as to relevance, or asks the reviewer if it is “on the right track”, so to speak, thereby refining search parameters. In other words the computer is being “trained” by a human reviewer to sort through a myriad of documents to determine which ones may be of relevance to the reviewer’s case.
Once the “training” and the predictions of the computer sufficiently coincide, the computer is then set free to examine all documents. It may produce a “yes” or “no” answer to relevance or it may assign a relevance “score”, which a human can then use to “prioritize” review.
Magistrate Peck highlights that the TAR process is far different from a “keyword” search, which may produce many documents with the keywords present but irrelevant to the case. Indeed, in a “post-Peck opinion” case, National Day Laborer v. U.S. Immigration and Customs Enforcement Agency, (USDC, SDNY (10 Civ. 3488 (SAS) (7/13/12)), Judge Shira A. Scheindlin, criticizes the government’s “keyword” approach in a case “...which concerns the largest FOIA [Freedom of Information Act] search in the history of ICE [Immigration and Customs Enforcement Agency] and an enormous search for DHS [Department of Homeland Security]and the FBI [Federal Bureau of Investigation] as well...” She opines that keyword searches were not in keeping with “...twenty-first century technologies to effectuate congressional intent.”
The opinion issued by Magistrate Peck is lengthy in that the protocol to be used in the TAR is appended as an exhibit. The opinion points out that the process, and an agreement as to its nature and use, is the important matter in determining the viability of TAR. Also, Magistrate Peck discusses a scenario in which, unlike the Da Silva Moore case, one party might object to the use of TAR - rhetorically asking what might be a viable alternative, if any. Should a court order TAR in order to respond to the opposing party’s document request? Also, his opinion discusses the lack of application of FRCP 702 and the United States Supreme Court’s decision in Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 113 S. Ct. 2786 (1993) to the TAR process. On April 26, 2012, U.S. District Judge Andrew L. Carter, Jr., affirmed Magistrate Peck’s Order regarding the use of TAR.
A court might order TAR under the given circumstances of a case, or it might not if the cost outweighed the perceived benefit. That concept, generally, might be embodied in FRCP 26, one of weighing the cost-benefit analysis of TAR in any given proceeding. Decisions may look to cooperation between and among counsel to a matter to develop a TAR protocol. As Magistrate Peck pointed out, the process is the key: may one review it and be satisfied that it is undertaken in good faith and is a viable attempt to secure fair discovery. Cost saving is not the sole determinant in the use of TAR.
Last year, the United States Tax Court sanctioned the use of TAR. In Dynamo Holdings Limited Partnership V Commissioner, 143 T.C. 9 (9/17/14) the Tax Court allowed the use of TAR so that the petitioner, responding to a document request by the Commissioner, could have some assurance that it was not disclosing privileged information. The Commissioner asserted TAR to be an “unproven technology” and that by turning over all data requested the petitioner would save time and expense and could protect itself through an agreement to preserve the right to assert privilege at a later time. The Court rejected the Commissioner’s arguments, citing the cost and time savings in responding to the government’s request. If the Commissioner were to be unhappy with the result the Court stated that “...he may file a motion to compel at that time.”
In 2014 several cases discussed the use of TAR: one case allowed it, reasoning that while it was not perfect discovery, it was the best alternative; another rejected a unilateral approach to the use of TAR; another remarked that TAR was better than human or “keyword” searching; another allowed a change in technique during discovery by granting the use of TAR; and another limited the scope of discovery and the attendant use of TAR.