By John Papageorge
For attorneys handling larger litigation matters, electronically stored information – or ESI – has changed the landscape on how discovery is conducted. The days of manually reviewing boxes of client documents in cold warehouses have taken a backseat to wading through thousands – and in some cases millions – of emails and other electronic records.
Attorneys now are faced with the monumental task of collecting, reviewing and producing their own client’s electronic documents while also reviewing the opposing side’s electronic documents. This can lead to uncomfortable conversations with clients regarding the significant cost of the process.
Supporters of predictive coding argue it makes the electronic discovery process less costly and less complicated. Predictive coding – a type of technology-assisted review or computer-assisted review – uses computers and algorithms to identify relevant and responsive documents in an automated manner. Unlike manual review, where the review is done by the most junior staff, predictive coding involves a more senior attorney or small team who review a “seed set” of documents for responsiveness.
The predictive coding system then applies the algorithms to identify properties of the seed set to automatically code the documents not reviewed by the attorneys. As the attorney team continues to code or identify additional responsive documents, the computer predicts the responsiveness of the universe of documents. Attorneys must review sample sets of documents coded by the computer and ultimately decide they have satisfied the requirements of Rule 26. In the end, the computer can identify thousands or even millions of responsive documents without the need for manual attorney review, saving clients thousands of dollars.
Predictive coding is gaining support in federal courts. In February 2012, United States Magistrate Judge Andrew J. Peck of United States District Court for the Southern District of New York, a leading authority on predictive coding, approved the use of predictive coding in Moore v. Publicis Groupe based on the following reasons: (1) the parties’ agreement, (2) the vast amount of electronically stored information (over three million documents), (3) the superiority of computer-assisted review to the available alternatives (i.e., linear manual review or keyword searches), (4) the need for cost effectiveness and proportionality under Rule 26, and (5) the transparent process proposed by one of the parties.
While Judge Peck cautioned that computer-assisted review is not appropriate in all cases, he did urge the bar to seriously consider using predictive coding in large-data-volume cases where it may save litigants a significant amount of legal fees for document review.
In April 2013, U.S. District Judge Robert L. Miller Jr. of the Northern District of Indiana issued a ruling in In re Biomet related to a discovery dispute involving keyword searches and predictive coding. Biomet produced millions of documents by initially using keyword searches to narrow the field of documents, followed by predictive coding to identify relevant documents to be produced. Biomet spent millions of dollars on electronic discovery. Plaintiffs objected to Biomet’s search method and argued that predictive coding should have been used from the outset. Plaintiffs wanted Biomet to start the discovery process over.
While Judge Miller stated that predictive coding from the outset might have unearthed additional documents, he ultimately rejected plaintiffs’ request to start over because Biomet, through the use of keyword searches and predictive coding, had satisfied the requirements of Rule 26 and the cost to start over outweighed any benefits of starting over. Judge Miller did tell plaintiffs that if they wanted documents produced via predictive coding only, they could pay the additional costs.
In May 2013, in Gordon v. Kaleida Health, an employment matter pending in the Western District of New York, the parties asked the court to resolve a discovery dispute involving approximately a quarter of a million electronic documents. For more than a year, the parties attempted, without success, to agree on how to achieve a cost-effective review of defendants’ voluminous emails using keyword search methodology. The court was dissatisfied with the lack of progress using keyword searches, and it pointed to predictive coding as another option.
After defendants decided to use predictive coding, the parties then fought over plaintiffs’ use of a conflicted consultant. Plaintiffs also took the position that the parties must negotiate a transparent protocol to guide the use of predictive coding software. Defendants, on the other hand, asserted that courts do not order parties in ESI discovery disputes to agree to specific protocols to facilitate computer-assisted review, based on the general rule that ESI production is within the sound discretion of the producing party. Because defendants ultimately agreed to meet and discuss the production using predictive coding, the court did not rule on the protocol dispute.
As caselaw develops on predictive coding, the issue of transparency related to predictive coding will be one to watch.
For larger litigation matters, predictive coding is here to stay, although the process is ever-changing. Many issues need to be resolved related to predictive coding, but cooperation and transparency will certainly take center stage. And as costs continue to escalate, courts will be faced with ongoing disputes over who pays for what.
Even after these disputes are resolved, parties must find the key documents that can be used to win the case. Maybe searching boxes of documents in a cold warehouse was not so bad after all.•
John Papageorge – [email protected] – is a partner with Taft Stettinius & Hollister LLP, practicing complex civil litigation with significant experience handling issues related to electronic discovery. He serves as the firm’s e-discovery practice contact. The opinions expressed are those of the author.