Blog

They’re at it again! How Anne Kershaw and Joe Howie are changing the world of document review.

Anne Kershaw and Joe Howie are bound and determined to make the world a better place … the e-discovery world at least.  For the past few years,  Anne, Joe and their colleagues at the eDiscovery Institute have set out to demonstrate that the process of human review in discovery is not only costly and inefficient, but it’s also fraught with inconsistency.

As an organization that is in the business of reviewing documents, you might think that DiscoverReady’s hidden agenda at this point would be to hire a real life member of the Sopranos to pay Anne and Joe a little visit. Nothing could be further from the truth.

Solving some of the inefficiency and expense of traditional document review is the reason DiscoverReady has had such tremendous success.  Even though our document review process is continuously evolving, we work hard to ensure it is a little (actually, we believe, a lot) better than the alternatives.  And, as part of our “commitment to better,” we’ve taken the results of the review of millions of documents and developed the i-Decision™ process, our intelligent automated review offering.  We’re pretty excited about this service and so are our clients.  Simply stated i-Decision™ is less expensive than human review and it’s also more accurate.

This past October, Anne and Joe once again pushed forward the discussion of human review vs. automation by publishing a survey related to some of the participants in the world of computer-generated decisions.  The survey is a wealth of information and reflects favorably not only on the background work that Anne and Joe performed, but on the thought that the participants put into their offerings and their responses.

“Predictive coding” vs.” definitive categorization”

A careful reading of the survey reveals that many of the participants are delivering what Anne and Joe refer to as “predictive coding.”  This means that in most instances documents are not “definitively categorized,” but instead they are “ranked” or “prioritized” or that there is “suggested” review coding.  In almost every case the participants’ workflow then calls for these rankings to be used as a means to determine review sequence or batching (either to place the documents most likely to be relevant first in the review queue, or to group similar documents together in order to achieve greater consistency or review speeds).

While we value any toolset or process that makes the review process more efficient or consistent (which many of the techniques described in the survey clearly can do), our take is that most predictive coding solutions are simply not very fulfilling.  When we designed the i-Decision process we outlined a number of key development criteria—e.g., the process had to be as accurate or more accurate than human review, the process had to be less expensive and more efficient than traditional review, etc.—but next to the requirement that the review had to be highly accurate, our most important goal was that the review had to be “complete.”

That is to say, our goal with i-Decision is not simply to use automation to predict the potential relevance or similarity of a particular document, our goal is to use automation as a complete review substitute.  We believe that the right answer for review automation is not to tell the attorney in what sequence to review the documents, but that the automation should allow the attorney to trust the automated decision-making process completely and to confidently eliminate the need to physically review a significant percentage of the documents.

Both “predictive coding” and “definitive categorization” have value and potentially have a role in a comprehensive review process.  But the distinction between “predictive coding” and “definitive categorization” is real and it’s important.

Predictive coding gives the attorney-team a map and a means of making their review more efficient.  It’s a review enhancement feature or an early case assessment function (the type of service performed by our Dynamic Data Analysis team).  At DiscoverReady, we value and use predictive coding for both of these functions.

But “definitive categorization,” the hallmark of our i-Decision automated review platform, doesn’t just “help” and leave the real work for later.  It gets the job done.

Maureen O'Neill