D4 eDiscovery Service Blog
Jan 30

By Cindy Courtney, Esq., VP, Discovery Engineering, General Counsel, D4

A Tuesday afternoon panel brought together Jim Wagner and Amy Hinzmann, two senior consultants from DiscoverReady, along with Marla Bergman from Goldman Sachs, Liz Hyon from Barclays, and Craig Cannon from Bank of America who gathered to consider the role of Predictive Coding (PC) in government investigations and regulatory matters. In particular, can PC be effective in responding to “produce all” requests by regulators? Marla Bergman from Goldman Sachs described an investigation in which the regulator asked whether Goldman would consider using PC, to which the response was “of course.” Ultimately, Goldman chose a vendor and negotiated a detailed protocol. Concededly, however, Goldman wasn’t really negotiating, because in the end they would do whatever the regulator required. Why lose sleep over a production that may be broad enough to include irrelevant material when the government can require the party to “produce all”?

The most intriguing views expressed by this panel concerned the detailed collaborative protocol negotiated in In re Actos, which we wrote about here and here. Craig Cannon believes that cases such as In re Actos are actually chilling the use of PC because the resulting broad disclosure of documents that may be completely irrelevant is discouraging parties who would otherwise embrace PC. His stance: Advocate for strictly defined parameters. If the regulator pushes back, don’t opt for Predictive Coding.

The regulator always has the option of taking everything and then using the PC tool themselves, but the target corporation need not be put in a position of sharing its thinking and strategy through implementation of an Actos-style protocol. Jim Wagner agreed that the cases are having a chilling effect on parties’ interest in PC. He refers to the Actos protocol’s requirement that required lawyers from the opposing sides to collaboratively review the training set as an exercise in “kumbaya.” However, what happens when disputes over individual documents are submitted to the courts under the protocol, causing a discovery stalemate? What starts out as a safeguard, or a check-and-balance, turns into an opportunity for one side to hold the process hostage.

Amy Hinzmann raised another critical post-Actos question: If we agree to an Actos-style protocol, how do we know that the collaborative review of the training set won’t result in a tactical battle of wills over documents that aren’t relevant, but that the receiving party wants to have? To protect their position in the process, the parties will want to assign senior lawyers to the collaborative review, but no one wants to take a senior lawyer out of circulation for what could turn out to be weeks.

The panel identified the following keys to success for using PC in a government investigation:

1. Planning and practice before use in a big matter;
2. Using PC on a data set with adequate richness;
3. Assuring a quality review of the training set;
4. Ability to successfully negotiate with opposing counsel or regulatory agency;
5. Effective claw-back agreement;
6. Staying open to opposing counsel actively participating;
7. Employing a quality privilege screening process;
8. Consider a “make available” production in an online review tool, either of the highest ranked documents or even the documents that are indeterminate;
9. Documentation of process and audit trail;
10. Defensible statistical sampling and validation model;
11. Mocking your 30(b)( 6) and pretrial conferences;
12. Understanding whether you are actually achieving a cost savings; and
13. Taking into account technology, consulting, project management and all forms of review.

Issues surrounding Predictive Coding, whether in regulatory or private matters, have a long way to go before they are ironed out.

Tags: , , , , , , , , , , ,

Leave a Reply


Connect with D4