In what appears to be the first decision to approve the use of predictive coding in a federal case, Magistrate Judge Andrew Peck approved a joint Discovery Protocol which would allow the defendant to use computer-assisted review, i.e., predictive coding, in a gender discrimination and Equal Pay Act action brought in the Southern District of New York. Moore v. Publicis Groupe, --- F. Supp.2d ----, 2012 WL 607412 (S.D.N.Y. Feb. 24, 2012). Magistrate Judge Peck specifically noted that his decision “now recognizes that computer-assisted review is an acceptable way to search for relevant ESI in appropriate cases.” Although Magistrate Peck held that the use of predictive coding was not subject to a Daubert analysis, he noted that he “would be interested in both the process used and the results.” The parties were considering predictive coding as there were approximately three million electronic documents from the agreed-upon custodians.
The discovery protocol provided that defendant would create a “seed set” that would be used to train the predictive coding software, through a combination of what was called “judgmental sampling” and relevance coding of the top 50 hits from keyword searches by senior attorneys. These documents would then be provided to the plaintiffs, with certain privileged documents withheld.
In addition, defendant proposed using 7 “iterative rounds” to fine tune the precision with which the predictive coding software would return relevant documents. Magistrate Judge Peck noted that additional iterative rounds would be run if necessary.
Judge Peck held that the use of predictive coding was appropriate for a number of reasons: (1) the parties agreed to its use; (2) there was a vast amount of ESI to be reviewed; (3) predictive coding was superior to other available alternatives; (4) the process would be cost effective; and (5) the process proposed by the defendant was transparent. He noted that the decision to use predictive coding would be “slightly more difficult when the producing party wants to use computer-assisted review and the requesting party objects.”
Magistrate Judge Peck ruminated on certain lessons for the future, including: (a) that it was unlikely that courts can determine or approve a party’s proposal as to when review and production can stop until the predictive coding software has been trained and the results are quality controlled; (b) targeted discovery focusing on the most likely relevant sources is a way to control discovery costs; (c) counsel should draw on the client’s knowledge of the opposing party’s records; and (d) it is desirable to have individuals familiar with predictive coding at hearings where discovery protocols are discussed.
Plaintiffs objected to Magistrate Judge Peck’s decision on a number of grounds, including, among others, that there was no evidentiary hearing on the reliability and accuracy of the predictive coding software defendant was proposing to use, that the proposed software would fail to capture up to 65% of the relevant documents, and that Magistrate Judge Peck failed to disclose his extrajudicial activities and interactions with both the defendant’s eDiscovery counsel and the maker of the predictive coding software being used. District Judge Andrew L. Carter must now decide whether Magistrate Judge Peck’s decision will stand.