Reprinted with permission of the Fulton County Daily Report

Special to the Daily Report  

When I was a junior associate, the partner for whom I worked made clear that while I may dislike document review, it provided job security because I knew the documents. The same is true today; the only difference is how we get to a manageable set to "know."

Long ago we ceased using entry-level associates to man large-scale document review. Since at least the late 1990s and through today, we tend to employ a three-tiered approach to document review: (1) electronic collection culled by keyword; (2) first-level review by vendors or in-house contract attorneys with coding; and (3) final review by firm counsel for production.

There certainly are variations on this basic process, but it is one that works well to reduce a potentially large volume of documents to a manageable collection from which key documents can be highlighted and tagged, and responsive documents produced.

Where predictive coding is helpful is to replace keyword searching and to reduce the time, but not eliminate the time, spent on steps two and three. This is particularly important in cases where the collection of documents is in the millions—well beyond the size of what any normal conference room could hold in bankers' boxes. Indeed, benefits of predictive coding may be felt most where relevant materials may be hard to identify by conventional means or may represent a very small portion of the culled set. Predictive coding is also likely to be used in combination with other tools to derive the most responsive set of documents.

Even with that benefit, the algorithms cannot ultimately replace the human element, particularly the lawyers' responsibility to know the key facts and documents in play and to ensure a full and complete production. Ultimately, lawyers will not be able to abdicate their role to machines. Associates play a key role in ensuring that the team (in large cases) or the lead attorney (in smaller cases) has all the facts at their disposal.

Additionally, predictive coding as of yet cannot, to my knowledge, address issues relating to multinational proceedings and multiple languages.

Recently when I visited my firm's Columbia, S.C., office, our document review center was conducting a large-scale review project all in French with attorneys all fluent in the language. Not only are there potential language barriers but also cultural ones that bear upon coding email in particular, where there is little or no way to understand whether a comment is made literally or in jest or is merely a colloquialism. For that, the human element to make those distinctions is absolutely necessary, and it's key for ensuring that the clients and lawyers fulfill their discovery responsibilities in litigation.