In In re Biomet M2a Magnum Hip Implant Products Liability Litigation, No. 12-MD-2391 (N.D. Ind. Aug. 21, 2013), the defendant used predictive coding to locate and produce responsive documents from a “universe” of 2.9 million documents.  The court rejected the plaintiffs’ motion to compel the defendant to produce the “seed set” of documents used to “train” the predictive coding software to identify relevant documents.  The defendant produced all of the documents used in the seed set that were determined to be relevant, but did not specifically identify them for the plaintiff, and did not produce any of the seed documents determined to be irrelevant.  As to the latter category, the court held that “irrelevant or privileged documents used to tell the algorithm what not to find” were “self-evident[ly]” not discoverable.  As to the relevant documents used in the seed set, the court held that “Rule 26(b)(1) doesn’t make such information discoverable,” primarily because “[the plaintiffs] want[] to know, not whether a document exists or where it is, but rather how [defendant] used [them] before disclosing them.”  The court nonetheless held that the defendant should “re-think its refusal” to identify the relevant documents in the seed set because “[a]n unexplained lack of cooperation in discovery can lead a court to question why the uncooperative party is hiding something[.]”