Using continuous active learning to efficiently respond to broad DOJ Civil Investigative Demand

Oct 4 2019

A client facing an expansive Civil Investigative Demand (CID) from the U.S. Department of Justice turned to eMerge to craft a response strategy that would meet stringent DOJ standards while remaining within tight budgetary limitations. 

The Challenges

• CID focused on topics central to client’s operation, implicating 2.5 million documents
• Tight response timetable coincided almost entirely with COVID-19 pandemic
• Official DOJ guidance regarding predictive coding still modeled after outmoded TAR 1.0 methodologies
• Litigation team separately needed to quickly identify and assess underlying issues reflected in document set

eMerge Solutions

eDiscovery Advocacy & Strategy

• Negotiated a protocol with the DOJ predicated on the use of continuous active learning (TAR 2.0) to define our outgoing production that met the DOJ’s heightened statistical and validation requirements for using predictive coding while requiring no client disclosure of either substantive coding information or examples of non-responsive documents
• Devised and implemented a mid-review workflow adjustment in response to unexpected client resource constraints due to the economic downturn caused by COVID-19, providing client with viable options for further reducing the costs of the review

Managed Review & Custom Technology-Assisted Review Workflows

• Facilitated an accelerated, highly efficient deep dive into the key factual issues that governed the core of the investigation via an active learning-powered review of just 4% of the documents in the population.
• Successfully redirected the focus of the review to new avenues of inquiry as the case team learned about critical events and decisions, and as more nuanced factual questions came to light, enabling the case to advise our client earlier in the process on how the DOJ investigation was likely to impact its business
• Used the results of the earlier issue-focused CAL review to triage predicted-responsive documents based on their proximity to the central issues of the case, allowing QC efforts and staffing to be adjusted to give more attention to the most important documents being produced, while saving costs on the remainder