News & Events

2012: A Document Review Odyssey – Federal court approves use of predictive coding over keyword searching

In a bellwether case allowing the use of technology to assist with document review, Judge Andrew Carter Jr. of the U.S. District Court for the Southern District of New York upheld Magistrate Judge Andrew Peck’s rulings approving the use of predictive coding in Da Silva Moore v. Publicis Groupe, 1:11-cv-1279-ALC-AJP. Judge Carter’s April 26, 2012 opinion can be found here. Judge Peck’s February 24, 2012 ruling can be found here.

Predictive coding, also called computer-assisted review (“CAR”), is the use of sophisticated software to code documents as relevant or non-relevant without having a human manually review each one. On a large scale, this can save hundreds of hours and tens of thousands of dollars. The process works by having a senior attorney manually review a small seed set. The software analyzes the text of the documents in that set and learns the differences between a document the attorney said was relevant and one she said was not. The review tool will then apply a sophisticated algorithm to the entire data set, and predict how each document should be coded. To make the process reliable and stable, the attorney will manually review some of the predictions. This review-predict-review cycle continues until the reviewer’s coding and the computer’s predictions are sufficiently similar (a 95% confidence interval was agreed to in Da Silva Moore). Once the system is stabilized, manual review is no longer necessary except for occasional quality control of small random data sets. This process allows a review team to code hundreds of thousands of documents in the time (read: money) it takes to manually review only a few thousand.

Da Silva Moore is such an important case because previously no court has formally approved the use of CAR. Having the blessing of the Southern District of New York, where renowned e-discovery judge Schira Scheindlin sits, is certain to increase the proliferance of predictive coding “in appropriate cases.” Judge Peck, reiterated by Judge Carter, noted what makes a case appropriate for CAR use: large scale; mutual involvement in the process; and transparency. The Court also noted the benefit of having the vendor present to the Court, what Judge Peck called “bring your geek to court day.”

Expect to hear from your outside counsel about the potential benefits of using predictive coding or computer-assisted review in large-scale document review cases. If not, discuss it with them. It will become the new normal, and could save you an incredible amount of money.

John H. Dollarhide