Balancing Work Product, Predictive Coding, Cooperation and Transparency in Bridgestone and Progressive

A pair of recent decisions emphasize the need for transparency, cooperation and openness in eDiscovery in general, and in particular where predictive coding is used. These cases are in conflict with the expectation that predictive coding may be an element of eDiscovery strategy, which is part of an overall case strategy which merits privilege protection under the work product doctrine.

In May, in Progressive Casualty v. Delaney the District of Nevada rejected Progressive's application to amend the operative ESI order to permit it to make use of predictive coding. The court made made particular mention of the fact that Progressive was "unwilling to engage in the type of cooperation and transparency that is needed for a predictive coding protocol to be accepted by the court or opposing counsel as a reasonable method to search for and produce responsive ESI."

In July, the Tennessee District court in Bridgestone Americas v. IBM permitted Bridgestone to utilize predictive coding on a set of documents already culled via search terms, noting however that "openness and transparency in what Plaintiff is doing will be of critical importance..."

These cases draw on the spirit of the Rule 26(f) conference, which is designed to promote cooperation among parties in eDiscovery. Courts have long expected litigants to approach discovery obligations in the spirit of cooperation and to develop collection, review and production strategies for discussion in the meet-and-confer. These recent predictive coding cases extend the idea of cooperation and create an expectation of disclosure and openness that cuts close to strategic areas that many litigants have long considered to be protected work product.

Litigants have often claimed that eDiscovery strategies including search terms and methodologies are protected by the work product privilege. EDiscovery plans including strategies and techniques for culling data reflect the mental impressions of the attorney, the argument goes. Requiring disclosure of those strategies and tactics would expose those mental impressions and inhibit zealous representation.

Predictive coding arguably falls into the same bucket as search terms and other elements of eDiscovery strategy. Counsel's decision to use predictive coding can be viewed as a strategic decision reflecting overall discovery strategy and intimate knowledge of the landscape of the collection gathered via witness and custodian interviews. The predictive coding process itself almost inevitably involves counsel's mental impressions of seed or training documents as well as decisions about what sorts of documents might or might not be responsive.

The argument for such work product protection is supported by the pre-eDiscovery case Sporck v. Peil, 759 F.2d 312 (3d Cir. 1985) in which the Third Circuit held that "the selection and compilation of documents by counsel...in preparation for pretrial discovery falls within the highly protected category of opinion work product."

Of course, a number of courts have rejected the claim that Sporck mandates protection of search terms. See, for example, Romero v. Allstate Ins. Co., 271 F.R.D. 96 (E.D. Pa. 2010) (search terms related to facts and did not fall under the protection of the work-product doctrine); FormFactor, Inc. v. Micro-Probe Inc., 2012 U.S. Dist. LEXIS 62233 (N.D. Cal. May 3, 2012) (rejecting plaintiff’s argument that disclosure of search terms would reveal privileged information); Apple Inc. v. Samsung Electronics Co., LTD., 2013 U.S. Dist. LEXIS 67085 (N.D. Cal. May 9, 2013) (third-party ordered to produce the search terms it used in responding to request for production, citing traditional principles of transparency and collaboration in discovery).

The recent decisions in Bridgestone and Progressive Casualty suggest that courts place predictive coding in the same category as search terms and other "fact work product" not entitled to privilege protection. These cases reflect a trend to require further transparency in the pursuit of cooperation, and the drive to openness has extended to the use of predictive coding.

Litigants who expect to make strategic use of predictive coding without disclosure may find themselves the subject of judicial disapproval, not because courts don't approve of the use of predictive coding technology but because they expect litigants to be transparent about its use. At a minimum these cases suggest that fine distinctions will need to be made between the strategic aspects of predictive coding's use and the purely factual ones.