Technology Assisted Review in Government Investigations: Where Are We Now?
eDiscovery comes in many shapes and sizes and some practice areas are naturally more subject to dealing with data intensive production requests than others. One such area is government investigations. Whether it is an investigation led by the DOJ, SEC, FTC or any of the other myriad Federal or state agencies, being on the receiving end of a government inquiry often means having to collect and make sense of an exorbitantly large amount of data.
--- Enter technology-assisted review (TAR) ---
TAR has been the greatest technological advance in cost-effective review of large data sets, but its acceptance by the legal profession has been slow. With that in mind, how do government agencies feel about the use of TAR in response to an investigation?
What is TAR?
TAR is a member of the family of eDiscovery analytics technologies. This family can be broken down into three basic categories: Structured, Conceptual and Predictive (TAR).
Structured Analytics includes basic essentials of an eDiscovery toolkit, such as email threading, near-duplicate identification, language identification and repeat content identification. These are quick and easy ways to eliminate redundant data. In fact, something like email threading is so commonly used it is essentially a default workflow component for many of our clients.
Conceptual Analytics uses an index to identify relationship patterns between words and concepts within a data set and understand the semantic meaning and context of terms. For example, users can group similar documents together, search for certain concepts within a data set and use a specific document to find others that are conceptually similar. While not used quite as commonly as structured analytics, conceptual analytics can provide quick insight into large data sets and help inform decision-making on which data requires linear review.
Finally we come to Predictive Analytics (TAR), an iterative process of teaching a review platform how to code documents. TAR is the most advanced technology available for reviewing large sets of data while minimizing the time and costs associated with linear review. TAR involves a set of human reviewers coding a “seed set” of documents and training a platform on what they are looking for until the platform can accurately predict, with a relatively high degree of certainty, how the human reviewers would code the remainder of the data set.
TAR is not a technology solely germane to the legal profession. In fact, you most likely use it on a daily basis. Websites and applications that recommend books, music, clothes, etc., based on prior selections are using the same algorithm-based learning process found in TAR. The online music platform Pandora, for example, uses your repeated thumbs up or down voting on songs to refine its algorithm. The more input you provide, the more the algorithm provides you with songs that more or less match your tastes. The same holds true for TAR in the discovery process.
Where are we now?
We have come a long way since TAR’s early days when it was something the legal profession approached with great trepidation. Nowadays courts widely approve its use and government agencies have largely embraced it. The DOJ Antitrust Division (Division), for example, issued an updated Model Second Request in November 2016 with revised instructions on the use of TAR. The instructions focus not on whether TAR should be used, but when and how to use it. In fact, the Division began inviting parties to use TAR in Second Request productions back in a 2014 initiative. According to an ABA March 2017 white paper by Tracy Greer, Senior Counsel-Electronic Discovery for the Division, the initiative was met with great success in the “majority of investigations,” which spurred the incorporation of specific TAR instructions into the November 2016 Model Second Request.
Other agencies similarly embrace TAR’s use. For example, the FTC’s Model Request contains instructions for the use of TAR, and the SEC will allow it subject to prior approval. Randy Lehner, a partner in the Chicago office of Kelley Drye & Warren, LLP focusing on regulatory investigations, government enforcement actions and related litigation, has seen a consistent theme with respect to how government agencies approach eDiscovery technology:
"In my experience, government agencies are willing to work with you over the scope and methods of a subpoena production. If you can demonstrate that your search and production process for the requested documents is reasonable given the issues and that you are open to an iterative process if the agency believes that your production needs to be expanded, a method like TAR should be acceptable."
Inventus has managed countless data productions in government investigations over the years, for clients ranging from Fortune 500 corporations to individuals. Our experience has shown us that various agencies welcome the use of TAR and are knowledgeable about the processes and tools available. The key to successfully using TAR is getting the agency’s acceptance early on by showing them you intend on using an established platform and detailing your proposed workflow, as well as following the respective agency’s guidelines where applicable. As long as the agency is satisfied with these aspects, you should encounter little resistance.