By Tom Groom, VP, Discovery Engineering
Computers are assistants to humans, consistent but not very smart without the action of the human being. You turn on your computer and it does nothing until you interact with it. Humans, on the other hand, are smart but not very consistent. What we need for dealing with the data deluge in eDiscovery is a prioritization system that leverages the efficacy of both human and computer. Seriously, who wouldn’t want that? Human review would still be conducted on the most relevant population—it would just be organized in priority order.
Predictive Coding/Technology Assisted Review (PC/TAR) provide the solution for “document prioritization” and are based on technologies that have been in use in other industries for years. While the end goal of Predictive Coding and Assisted Review is the same—namely, prioritizing document collections for eDiscovery workflows—there are fundamental differences in how the two approaches achieve prioritization and when they can be used in the discovery lifecycle.
The primary differences between Predictive Coding and Assisted Review fall into these general categories; Faster Results, Flexibility in Workflow, and Support for Defensibility.
Faster Results – You may get close to the same end results with either approach, but you will get there sooner with Predictive Coding. With Assisted Review you will need to manually tell the software what is relevant until it is “good enough” (normally after review of 10,000 to 15,000 documents taking 10 to 15 days), but with Predictive Coding, the platform actively determines the difference between “relevant” and “non-relevant” and lets you know when you’ve achieved your goal (normally after review of 2,500 to 3,500 documents taking 2 to 4 days).
Flexibility in Workflow – Many Predictive Coding systems run external to the review system and therefore enable workflows before the review process, which can significantly reduce overall costs. Some Predictive Coding systems can also be integrated with Early Data Analysis and processing tools to make a more complete “data management” solution. Assisted Review systems are embedded in the review platform and are focused solely on assisting review.
Support for Defensibility – Defensibility for both approaches is a function of the workflow surrounding the technology. Predictive Coding tools tend to provide more support to aid that process. It not only gives you your results: A, B and C, but also tells you how you got A, B and C and what parameters were used to obtain those results.
You can learn more about Predictive Coding and Assisted Review from the links below: