The recent predictive coding wars have led to a separate but related battle over how much information is appropriate to share—or to force a party to share—as part of the seed set and training process. One recent example is the much-litigated protocol for production of electronically stored information (ESI) in Moore v. Publicis Groupe,1 in which a federal court concluded that computer-assisted review was an appropriate means of determining relevant and discoverable documents for production. Addressing the e-discovery process more broadly, the court noted that, “[e]lectronic discovery requires cooperation between opposing counsel and transparency in all aspect of preservation and production of ESI.”2
“Cooperation” has been an e-discovery buzzword for years, originally intended as a way to encourage parties to stop engaging in e-discovery gamesmanship.3 Over the ensuing years, the idea of cooperation has, perhaps not surprisingly, turned from a shield into a sword in the hands of some parties—and judges. “Forced cooperation,” in the form of directing a party to turn over something in addition to documents it has deemed responsive to a document request—such as search terms or information related to predictive coding seeding—is not only a possible oxymoron, but also a dangerous dance on the edge of violating work product protection.
The question regarding what to share as part of the predictive coding process may not yet be ripe, as we are still grappling with the issue in everyday, garden variety e-discovery that involves keywords and other search terms. Indeed, in such matters, there has been a recent, unsettling trend by judges to order the production of search terms as part of an effort to settle discovery disputes. Over two years ago, the authors of an article in this publication cautioned judges to order the disclosure of search terms with care.4 Instead of heeding the counsel provided in that article, judges have continued to order the disclosure of search terms in a series of decisions that appear to discount or misunderstand the protected nature of key aspects of the e-discovery process. In this article, we review a recent decision that exemplifies this trend.
In a Sept. 11, 2013 opinion, a Nebraska federal judge presiding in American Home Assurance v. Greater Omaha Packing, ordered the defendant (GOPAC) to disclose the search terms it used to comply with co-plaintiff Cargill’s request for emails and other ESI.5 After receiving only 25 emails from GOPAC in response to its initial document production request, Cargill sought an order from the court to compel GOPAC to produce withheld emails, based on the premise that the small number of emails produced “evidences a lack of diligence in production.”6 The court denied Cargill’s motion, finding that it “cannot compel the production of information that does not exist.”7 GOPAC argued that it had already produced all ESI generated as a result of its searches in the months preceding the motion to compel.8 Additionally, the court noted that “GOPAC has also offered to search available sources using search terms provided by Cargill, but Cargill has refused to supply any additional terms.”9
Despite denying its motion, the court did not leave Cargill empty-handed; it ordered GOPAC to share with Cargill the search terms it used to search for relevant emails and ESI. Reaching this decision, the court wrote:
Discovery on this matter has been ongoing since July of 2012. It is unclear to the Court why ESI that has presumably been in GOPAC’s possession since the start of discovery has not been fully produced. To provide Cargill an adequate opportunity to contest discovery of ESI, the Court will order GOPAC to disclose the sources it has searched or intends to search and, for each source, the search terms used.10
Where’s the Reliance?
Cargill’s motion to compel notes that after “a telephone conversation and several letters, GOPAC has stated that it would search for emails and other responsive electronic data, and propose search terms, but Cargill has seen no action or supplemental production to date.”11 The court, though, does not mention this purported arrangement in its decision, so it is unclear if the order to produce search terms relies on the arrangement or was fashioned from whole cloth.
Whether or not the court relied on an existing arrangement between the parties goes to the heart of how courts should exercise their power within the discovery stage of litigation. Even without a previous understanding between the parties, the American Home court had the power to fashion a remedy it deemed appropriate to resolve the discovery dispute between the parties. Since the advent of e-discovery, judges have been asked, with increasing frequency, to step into the breach of a document production dispute and direct an outcome. In these instances, though, a court should tread carefully to ensure that it does not establish additional discovery burdens for a party whose actions have been reasonable under the circumstances, and does not order a party to produce documents or other information that may be protected under the attorney work product doctrine.
A Growing Trend: Disclosure
Whether American Home was decided based on GOPAC’s prior representation may be in question, but there is, indeed, an increased practice of court-ordered disclosure of search terms. Over the past several years, federal judges have not only ordered the disclosure of search terms, but have also opined that these search terms are not protected by privilege. A few examples include the following:
• In Romero v. Allstate Insurance, an Eastern District of Pennsylvania judge concluded that search terms were not subject to work product protection because they went to the underlying facts of what documents were responsive and did not delve into the thought processes of attorneys.12 The court ordered that the parties confer and “come to some agreement on the search terms that defendants intend to use, the custodians they intend to search, the date ranges for their new searches, and any other essential details about the search methodology they intend to implement for the production of electronically stored information concerning the Release.”
• In Formfactor v. Micro-Probe, a federal judge in the Northern District of California ordered a plaintiff to turn over search terms that were used to conduct an electronic search.13 Referencing Romero as support, the court wrote that “[s]uch information is not subject to any work product protection because it goes to the underlying facts of what documents are responsive to defendants’ document request, rather than the thought processes of plaintiff’s counsel.”
• In Apple v. Samsung Electronics, a federal judge ordered Google to comply with Apple’s request for search terms despite the fact that Google was not a party to the litigation between Apple and Samsung.14 In that case, also in the Northern District of California’s federal court, the judge found that third-party status did not confer any additional protection to compliance with a discovery order. Relying on Formfactor for the inapplicability of the “work-product immunity doctrine,” the court ordered Google to turn over to Apple the search terms it used in response to Apple’s request for the production of certain documents.
An Argument for Protection
In their attempts to draw the line between discoverable and nondiscoverable information, the above cases run the risk of over-simplifying the processes that parties use to comply with discovery requests. Lawyers are often responsible for sifting through thousands, even millions of documents and available information to comply with these requests. And, despite the changes resulting from an increased availability of electronic information, the search process is more complex than simply the generation and re-generation of words that appear to be relevant. Rather, the process is one requiring lawyers (or non-lawyers at the direction of lawyers) to search information while also developing (and redeveloping) a theory of the case, along with possible legal strategies. The process, which describes the necessary steps in even the most basic search to comply with a discovery request, is captured in FRCP 26(b)(3)(B), outlining the attorney work-product doctrine and stating that the court “must protect against disclosure of the mental impressions, conclusions, opinions, or legal theories of a party’s attorney or other representative concerning the litigation.”15
In the prior article on this topic noted above, the authors discussed a concern that some courts were considering the ease with which a party could share its search terms as justification that a party must disclose them.16 Over two years later, with more decisions going down this road, we echo this concern. Though computer technology has changed the process of searching for and through documents, it has yet to change the way that attorneys interact with those documents, along with the information contained within.
This point—that the technological ease with which a search term can be applied does not diminish the extent to which it represents an attorney’s mental impressions or opinions—takes on even greater significance as we consider more recent changes to the nature of electronic discovery. Increasingly, parties engaged in litigation involving vast quantities of documents are employing predictive coding to determine which documents are relevant to an opponent’s discovery request. As mentioned above, a small number of courts have ruled on discovery issues related to predictive coding. It would not be unfathomable to conclude that many judges will consider the relationship between generating search terms and attorney work-product to inform their perspectives on whether the disclosure of information related to predictive coding warrants work-product protection.
The Right Role for Courts
Courts are essential players in the discovery process. They help to ensure that parties comply with discovery agreements, as well as procedural rules. Above all else, they ensure that the process is fair. The moments when communication and compliance between the parties have broken down are when courts face the greatest challenge. And in these moments judges have the ability to significantly affect the course of litigation. When judges make their decisions based on previous agreements reached by the parties, they reinforce the twin propositions that the discovery process is best managed by the parties themselves, and that courts are at their best as infrequent arbiters of the process.
The disclosure of search terms exemplifies a judicial remedy that should be employed only when the parties have previously agreed to it. In those instances when such disclosure was not previously contemplated by the parties, we hope that judges will act with restraint, recognizing that a search term is more than a collection of words, rather, the culmination of an attorney’s interaction with the facts of the case.
H. Christopher Boehning and Daniel J. Toal are litigation partners at Paul, Weiss, Rifkind, Wharton & Garrison. Charles J. Hamilton, a law clerk, and Ross M. Gotler, e-discovery counsel, assisted in the preparation of this article.
1. 287 F.R.D. 182, 193 (S.D.N.Y. 2012).
2. Id. at 191 (quoting William A. Gross Const. Associates v. American Mfrs. Mut. Ins., 256 F.R.D. 134, 136 (S.D.N.Y. 2009)).
3. See “The Sedona Conference Proclamation,” available at http://www.thesedonaconference.org.
4. See David J. Kessler, Robert D. Owen and Emily Johnston, “Search Terms Are More Than Mere Words,” NYLJ (March 21, 2011).
5. No. 8:11-cv-270, 2013 WL 4875997 (D. Neb. Sept. 11, 2013).
6. Id. at *6.
10. American Home, 2012 WL 2061941, at *6.
11. Pl.’s Mot. to Compel (July 18, 2013) (emphasis added).
12. 271 F.R.D. 96 (E.D. Pa. 2010).
13. No. 10-cv-03095 PJH, 2012 WL 1575093 (N.D. Cal. May 3, 2012).
14. No. 12-cv-0630-LHK, 2013 WL 1942163 (N.D. Cal. May 9, 2013).
15. Fed. R. Civ. P. 26(b)(3)(B).
16. See David J. Kessler, Robert D. Owen and Emily Johnston, “Search Terms Are More Than Mere Words,” NYLJ (March 21, 2011).