Time again for predictions. No, not about “the big game.” Vegas has that covered. I mean predictions about the future of e-discovery in 2013.
But first, a look back at my not-entirely-fearless 2012 predictions. About this time last year, I suggested that 2012 would be the year that predictive coding arrived, not just in chatter or sales pitches, but in substance. If you have kept up with headline cases such as Da Silva Moore, Kleen, Global Aerospace or Actos, you would probably agree — I got that one right!
So where do we go in 2013? Well, first, there are some unanswered questions from January of 2012. I’ll restate them below to save you the time of thumbing through old blogs. One year ago, I asked:
Does this mean that litigants finally will be required to disclose their discovery practices?
Like it or not, the resounding answer in Actos, Da Silva Moore and other matters is, to an unprecedented, extent — YES. Although the processes laid out in these matters likely will undergo significant refinement in the coming months and years, it’s safe to say the genie is out of the bottle when it comes to increased transparency of discovery processes. Does this mean that a litigant must share ALL its secrets with the other side? No, but it’s clear that the proverbial kimono is more open than closed.
How good does a predictive-coding tool have to be to be defensible? If manual review is not much better than the flip of a coin, which some commentators and studies have suggested (erroneously I believe), then is a predictive-coding tool that’s right more than half the time good enough?
Right idea, wrong question. The question we should be asking today is not whether predictive coding is more accurate than human review, but whether predictive coding is more accurate than keyword search.
Why should we be comparing predictive coding to keyword search? Look at the published cases and protocols. Look at the battleground set in Kleen. Predictive coding’s most common use today is as a culling tool, in lieu of keyword searches, to more effectively find likely relevant files and to eliminate or reduce the volume of irrelevant data given to humans to review. In every major case published to date, there is no contemplation that the data with high “relevance scores” will be produced without human review.
Trust me when I say there’s no one with more at stake on this topic — the use of predictive coding as a human review replacement — than DiscoverReady. Starting in 2007, we undertook painstaking research and analysis, as well as substantial investment, all in support of our i-Decision® predictive coding process, which is designed as a “post-culling” complete first-pass review substitute. During this intensive R&D process, we demonstrated that automated review — when used thoughtfully and reasonably — is at least as good as, and in most cases materially more accurate and consistent than, human review. In their landmark article, Maura Grossman and Gordon Cormack agreed. Any suggestion to the contrary from the staffing world or others is at best suspicious. It reminds me of a certain e-discovery company’s self-serving and now-infamous last-gasp public pronouncement that native file review was DOA and that only .pdf review would serve the legal community’s interests.
Fortunately for us (and our clients), we have been able to marry the best of both worlds — basic predictive coding for the culling process, and our advanced i-Decision process for complete human review replacement. We call the combination Predictive Coding Plus™ and based upon our results to date, we feel very good about our position in the predictive coding world.
We know that the leading legal commentators want sound statistical sampling. But what does this mean? Who is going to design the protocols? Who is going to execute them? Measure the results? Document the process?
Although the jury is out on this one, based on first-hand experience it appears that the majority of this heavy lifting will take place by vendors and consultants, plus a key contact within the law firm, each of whom has developed substantial expertise in the area.
Most importantly, will predictive coding actually save clients money?
The answer to this is a clear, concise and easy … “maybe.”
[To be continued …]