Predictive Coding in E-Discovery Addressed in Federal Court

Predictive Coding in E-DiscoveryA recent hearing before Magistrate Judge Andrew J. Peck of the U.S. District Court for the Southern District of New York shed light on the potential complexities of using predictive coding in e-discovery.  This ruling appears to confirm what many e-discovery companies have been advising for some time – that predictive coding offers an alternative to manual review-only of large document sets, but the devil is in the details when it comes to ensuring accurate results and fairness to all parties.

Da Silva Moore v. Publicis Groupe et al.

The hearing came in a Title VII gender discrimination action alleging discriminatory practices against women.  The parties agreed to use predictive coding in e-discovery, but couldn’t agree on several details regarding its application.   The methodologies for training the predictive coding software, and verifying the results, were worked out in court through a wide-ranging discussion led by Judge Peck.  Each party’s expert vendor participated in the hearing, which included a disagreement between them as to how the predictive coding software actually works – complete with references to the software’s patents.


Judge Peck ruled for compromise on several protocol details, but largely accepted the producing party’s (defendant’s) proposal.  The Judge, however, cautioned that if the protocol did not produce acceptable results, the defendant might be required to incur the costs of repeating the predictive coding review, under a modified protocol.  This appears to be in line with previous comments from Judge Peck, where he has emphasized the accuracy of results more than the process.

Judge Peck suggested that he might issue a written opinion to help other litigants, and analysts are hoping he’ll provide general guidelines for establishing accuracy, transparency, and fairness in protocols for using predictive coding in e-discovery.  The Judge did offer several interesting comments, including an endorsement of predictive coding in e-discovery, generally:

“It certainly works better than most of the alternatives, if not all of the alternatives.  So the idea is not to make this perfect, it’s not going to be perfect. The idea is to make it significantly better than the alternative without nearly as much cost.”

The Judge also advised the parties to track their costs for a potential ruling on proportionality.

The anticipated written opinion from Judge Peck should provide more insight into the view from the bench on predictive coding in e-discovery.  What do you think about Judge Peck and his opinion on the use of predictive coding in e-discovery? Let us know in the comments below!

Send to a friend