By: Tom Barce
It is time to be honest about predictive coding. It is reasonable, it is defensible, it is better than manual review, it is applicable to a broad range of cases, and it has been around long enough for attorneys and even judges to accept and adopt the workflow. However, predictive coding is not just a technology. It is a process that only works well when practitioners engage the right subject matter experts, dedicate the right lawyers with substantive case knowledge, follow a good project plan, adapt to the technology, and are able to adjust expectations as unforeseen variables arise throughout the course of an imperfect process.
Wait a minute! You might be saying right now, “How could predictive coding be imperfect when my vendor told me that predictive coding technology searches and tags every document?” The answer is simple but the solution is multifaceted. Human language is not the same as bullets or pills. The natural process that a variety of individuals go through to generate content has innumerable, unpredictable variables. Through a predictive coding process, statistical sampling can ferret out the appropriate content, bring reliability and define the language that matters, but it will be different every time. It will be subject to more environmental factors than manufacturing, where statistical validation has the most notoriety and trust.
Furthermore, the majority of attorneys are not mathematicians or statisticians. At a recent CLE I gave at a major law firm regarding predictive coding, 1 out of approximately 20 attorneys was actually an engineer who had a first-hand appreciation for the mathematics behind predictive coding. Quite frankly, the majority of attorneys do not care about the math. They strive each day to balance the competing priorities surrounding the claims and defenses of each matter, rigorous representation of client interests, fulfillment of ethical obligations, and financial pressures that plague every legal budget in the wake of masses of electronic information.
Considering all of the above, attorneys need to embrace predictive coding from a perspective they are akin to, which has more to do with the substance of the information they have, relative to the merits of the case and the objectives of discovery. One can leverage the key elements of a predictive coding workflow to accomplish this. These elements include extensive use of content analysis throughout a predictive coding workflow. Attorneys have spent years hypothesizing about the content that might be within the corpus of their client’s documents and trying to prove their theories through keywords. For several years now, rich analytical tools such as concept clustering, combined with bibliographic metadata, reveal the content of a document universe to the stakeholders on broad or detailed levels. Instead of replacing the keyword guessing game with the mathematical guessing game, legal teams can become much more informed about the results of predictive coding and gain even greater advantages in their case by leveraging these tools.
Please stay tuned for future posts discussing the use of data analytics and predictive coding to increase the value in technology solutions to support your practice and meet business objectives.