It's hard to find anyone in the e-discovery field these days who's not talking about predictive coding, an artificial intelligence technique for finding relevant electronic documents that proponents say is a faster and cheaper approach than typical keyword searches and manual review.
If predictive coding is used correctly, a senior lawyer or small e-discovery team will need to review only a few thousand documents to train a computer to identify similarly relevant documents.
But for all its potential, many lawyers have been slow to embrace predictive coding. Some say the bench has sent mixed messages about its admissibility. But attorneys also may be dragging their feet because the technology is relatively new, at least as it's applied to legal discovery.
"It's not the courts who are hanging up predictive coding, it's the lawyers and the clients," says Ralph C. Losey, partner and national e-discovery counsel at Jackson Lewis. "Only the most sophisticated clients are asking about it and are saying to the lawyers, 'Hey, can't we save some money that way?' The rest of the bar is just starting to hear about it."
Losey estimates that, "all other things being equal," predictive coding could reduce discovery costs by 50 to 70 percent compared with keyword-linear review.
The technology underlying predictive coding is not new. Anyone who has trained a spam filter to reject certain types of email, or allowed Amazon.com to suggest books and CD titles to buy, has seen predictive coding in action. When it's applied to e-discovery, lawyers review a sample of a few thousand documents to train the computer to look for key issues; after that, predictive coding tools employ a variety of algorithms to find other relevant documents.
But the buzz about predictive coding has reached the bench. U.S. Magistrate Judge Andrew J. Peck of Manhattan became the first judge to publicly approve of predictive coding, or computer-assisted review, in the context of an employment dispute. "Counsel no longer have to worry," Peck wrote in February 2012, "about being the 'first' or 'guinea pig' for judicial acceptance of computer-assisted review." (Moore v. Publicis Grouppe
, 287 F.R.D. 182, 193 (S.D.N.Y. 2012).)
Since then, in Delaware's Court of Chancery Vice Chancellor J. Travis Laster issued a somewhat surprising bench ruling in October, mandating that both parties use predictive coding, and even share a common vendor. Laster ruled, "I would like you all, if you do not want to use predictive coding ... to show cause why this is not a case where predictive coding is the way to go." (EORHB, Inc., v. HOA Holdings, LLC,
No. 7409-VCL (Del. Ch. Ct. Oct. 15, 2012).)
This year, a federal court in California ruled that the use of predictive coding was "reasonable under the circumstances." (Gabriel Technologies Corp. v. Qualcomm Inc.,
2013 WL 410103 (S.D. Cal.).)
Some e-discovery experts say that other judges have insisted that any party using predictive coding must provide a much more transparent accounting of its discovery process than it would for using keywords and manual searches. "If that remains the case, that may have a chilling effect on the use of predictive coding," says Jeffrey Fowler, a partner in O'Melveny & Myers's Los Angeles office and one of the founders of the firm's e-discovery and document retention practice.
Meanwhile, e-discovery lawyers say many attorneys are looking for more than a few one-off rulings from the bench in favor of predictive coding before they throw in with the technology. Keyword searches have been around since the 1980s, but the approach didn't come into its own through judicial approval. Rather, enough lawyers started using it and over time it became a de facto standard. Predictive coding may have to go through a similar evolution over the next three to five years.