Because so many documents are stored nowadays in electronic form, the expense of electronic discovery has become a major concern in litigation. To address this concern, vendors have developed technologies to search through large pools of electronically stored documents for material responsive to e-discovery requests efficiently and expeditiously. One such new technology is “predictive coding” – a technique that uses computers to assist with review of large volumes of documents.
The United States Tax Court recently approved the use of predictive coding to respond to a large document request by the IRS. Dynamo Holdings, Ltd. v. Commissioner, 143 T.C. 9 (September 17, 2014).
The IRS alleged that various transfers (characterized as loans) from one entity (Beekman) to an affiliate (Dynamo) were disguised gifts to Dynamo’s individual owners. To substantiate its claim, the IRS asked the entities to produce electronic information stored on backup tapes. The entities objected to the cost of manually reviewing each document on the tapes, and petitioned the court to use “predictive coding” to efficiently and economically identify the non-privileged material responsive to the IRS’s request. The IRS opposed use of predictive coding as an “unproven technology,” and demanded manual review. According to the testimony of the electronic discovery expert retained by the petitioners, the difference in the time and expense between the alternative approaches was staggering – 200,000 to 400,000 documents subject to review at a cost of $80,000 to $85,000 using predictive coding versus 3.5 million to 7 million documents subject to review at a cost of $500,000 to $550,000 under the manual approach advocated by the IRS.
After providing a brief overview of predictive coding technology, the Tax Court rejected the view that predictive coding was an unproven technology. Citing opinions by several federal district courts, the Tax Court concluded that predictive coding has gained wide acceptance as a reliable tool to reduce the burdens of electronic discovery (tangentially, the Court also noted that a form of predictive coding is used by email applications to filter out unwanted spam).
The Tax Court’s conclusion relied heavily on testimony from the petitioners’ e-discovery technology expert, James Scarazzo (interestingly, the Court did not share the testimony of the IRS’s expert). Mr. Scarazzo opined that predictive coding was a superior technology because it eliminates (or least minimizes) human error and expedites review. As part of his mandate, Mr. Scarazzo reviewed the backup tapes targeted by the IRS, and compared the two approaches (predictive coding vs. manual review) in terms of time and cost. In doing so, Mr. Scarazzo assumed that the parties would use predictive coding to search the data using specific search criteria while the manual approach would not use search criteria. Mr. Scarazzo concluded that predictive coding would “minimize review time and expense and ultimately result in a focused set of information germane to the matter.”
Given that the Federal Rules governing discovery are to “be construed to secure the just, speedy, and inexpensive determination of every case,” the Tax Court sensibly authorized the use of predictive coding to respond to the IRS’s discovery request.
Related article: The Future of Predictive Coding - Rise of the Evidentiary Expert?
Have you had an opportunity to use predictive coding in litigation? If so, did you encounter any opposition from opposing counsel? Based on your experiences, do you feel that predictive coding is reliable?