John Rosenthal Quoted in Law360 Article "5 Predictive Coding Pitfalls to Avoid"

Winston & Strawn eDiscovery and information management chair John Rosenthal was quoted in the Law360 article “5 Predictive Coding Pitfalls to Avoid” on February 17, 2015.

The article reviews the eDiscovery technology known as predictive coding that “learns” which documents are relevant to a case and which are not, potentially cutting hundreds of hours and thousands of dollars from complex discovery review projects.

According to Mr. Rosenthal, one downside of using predictive coding technology to review documents is the technology’s struggle with distinguishing among multiple or overlapping legal issues that may have fuzzy lines. While predictive coding can be highly accurate in evaluating and connecting themes and grouping data within a set of clear parameters, testing on the tool’s ability to sort privileged documents from unprivileged ones has shown mixed results.

“At the end of the day, it’s going to give you back buckets of relevant and nonrelevant,” he said. “But privilege is a subtle thing, and it’s very difficult to train the engine for that.”

Mr. Rosenthal, who refers to himself as a predictive coding “realist,” said lawyers should know that vendor estimates on time and cost savings typically come from straightforward or single-issue eDiscovery efforts.

Predictive coding “just isn’t very good at parsing out different issues in a case,” he said. “You have to do different training for different issues, and that can become costly and eat into the savings.”