U.S. Magistrate Judge Andrew Peck’s declaration that computer-assisted review is “acceptable in appropriate cases” may have helped change the electronic discovery landscape forever. Prior to Judge Peck’s 2012 order in Da Silva Moore v. Publicis Groupe, there were no known cases specifically addressing the use of computer-assisted review (aka predictive coding technology). Since then, at least seven different courts have taken up the issue of predictive coding technology and when viewed collectively, the cases signify a trend toward continued judicial interest. For example, in October 2012, a Delaware Chancery Court Judge stunned many in the legal community with what appeared to be a sua sponte order in EORHB, Inc., et al v. HOA Holdings, LLC, when he asked the parties to show cause as to why they should not use predictive coding technology:
The lure of faster, cheaper, and more accurate document review combined with at least some judicial support is fueling predictive coding momentum. Despite the momentum, most remain on the sidelines because they are not quite sure how or when predictive coding technology should be used or how it is different from other tools in the litigator’s technology toolbelt™. Instead, they are taking a “wait and see approach” because the technology is still evolving and the options are not always clear. For example, following Judge Peck’s Da Silva Moore opinion, vendors continue to race to market at breakneck speed with their own versions of predictive coding technology. However, these technology solutions are not created equally or priced the same, and even the best technologies will not yield accurate results if they are used improperly. This kind of uncertainty is common with any evolving technology, but it creates something most litigators don’t like – unnecessary risk.
The Perfect World
In a perfect world, the risk-averse attorney could propose the use of predictive coding technology with no objection or input from the other side. In the real world of litigation, that may not always happen. For example, in Global Aerospace Inc., et al, v. Landow Aviation, L.P. dba Dulles Jet Center, defendants proposed using predictive coding technology to “retrieve potentially relevant documents from a massive collection of electronically stored information.” The judge granted defendants’ request over plaintiffs’ objection, but noted that the parties could object to the “completeness or the contents of the production or the ongoing use of predictive coding technology.” Fortunately for defendants, discovery between the parties continued with little fanfare.
The risk-averse attorney may look at Global Aerospace and shudder at the thought of defending the predictive coding methodology used. The fact that the protocol followed by defendants in Da Silva Moore was met with strenuous objection by plaintiffs and led to lengthy hearings and discovery disputes only compounds the fear of something going wrong. Further stoking those fears are recent observations in the ongoing In Re: Biomet M2a Magnum Hip Implant Products Liability Litigation (MDL 2391) case. In that matter, some outside observers (an academic and an attorney, both well versed in the field of predictive coding), suggested that some of the statistical calculations guiding the judge’s proportionality analysis may have been incorrect or misunderstood. If the observation is correct, then the implication is that the judge could have relied on more accurate information and required defendants to redo or supplement their productions. Instead, the judge found defendants satisfied their discovery obligation even though they may have excluded about half the relevant documents through keyword culling before predictive coding technology was used.
Do Lawyers Need to Be Statisticians?
The Biomet case spotlights the importance of statistics in predictive coding. Statistics are the backbone of any good predictive coding protocol and a common joke among lawyers is that they attended law school because they were bad at math. Not surprisingly, attorneys are beginning to believe that finding the right statistics expert is as important as choosing the right solution. The good news is that the risk of hiring the wrong expert, choosing the wrong predictive coding solution, or using a particular solution improperly can be mitigated by doing your homework. (Matthew D. Nelson, Choosing the Right Predictive Coding Software, (Chapter 7: Predictive Coding for Dummies, John Wiley & Sons, August, 2012).
For example, the most advanced tools automate complex statistical calculations that represent the most complicated, misunderstood, and riskiest step in the predictive coding process. If statistical calculations are not automated by the predictive coding tool, then the risk of making statistical errors is introduced every time a human being is required to perform a statistical analysis. On the other hand, if complicated statistical calculations are automated by the predictive coding tool, the need to hire an expert in statistics for every case is reduced or eliminated. The problem is that most lawyers don’t like homework and those who do, may not have the time. That is why in part two of this article, I will share the top three use cases for risk-averse and time-challenged attorneys who are interested in dipping their toes into predictive coding waters without taking on too much risk.