The use of predictive coding to help identify responsive discovery is a hybrid of science and the practice of law. Attorneys have to understand the merits of the case, the type of electronically stored information that can support the claims and defenses, and how to use the technology to find responsive discovery.
US District Judge Jon Tigar respected these complex issues in a discovery dispute where the Plaintiffs sought a rolling production of ESI while the Defendants were refining their technology assisted review (TAR) model. The Plaintiffs wanted ESI to support their condition certification motion. The Defendants [understandably] countered the timeframe for such a production was unrealistic. See, Rabin v. Pricewaterhousecoopers LLP, No. 16-cv-02276-JST, 2017 U.S. Dist. LEXIS 125404, at *2-3 (N.D. Cal. Aug. 8, 2017).
The Defendant’s ESI production workflow clearly used a predictive coding technology. The methodology used linguistic and statistical modeling on a set of documents narrowed by search terms in order to identify responsive ESI. Rabin, at *2 footnote 1.
The Court held that the Plaintiffs were entitled to discovery that supported their motion, but the issue remained, when could the Defendants begin a rolling production of ESI? Rabin, at *4.
Defendants argued that they could not start producing documents until December 2017. The Court likely raised an eyebrow at this projection, noting that the parties were near an agreement on search terms and that the TAR methodology would take weeks, not months, to identify responsive ESI, even while validating the TAR models during review. Rabin, at *5.
The Court acknowledged that the TAR methodology required attorneys to review a few thousand documents. That being said, the Court observed, “[W]ith appropriate resources devoted to this matter, there is no reason why that task cannot be completed within the next few weeks.” Id. Based on the above, Judge Tigar ordered the Defendants to begin production in August if possible, but not later than September 1, 2017. Rabin, at *6.
Bow Tie Thoughts
Predictive coding makes responding to discovery requests easier. There are multiple ways to use predictive coding. One option is to first review ESI based upon searches constructed for specific requests for production. Predictive coding models can be created to identify similar ESI based on specific coding paradigms of reviewed discovery, thus finding ESI that were not “hits” to searches, but are responsive to requests.
Another option is using predictive coding coupled with searches designed upon specific requests for production, thus focusing review on ESI likely to be responsive, opposed to all hits from search terms.
There are many other ways to leverage the advantages of predictive coding in responding to discovery requests. If you would like to learn more, I am moderating a webinar for Everlaw on November 1, 2017, at 1000am PST/100pm EST entitled The Guilt-Free Catch-Up on Predictive Coding and AI in eDiscovery. This will be an informative discussion on artificial intelligence in litigation through the use of machine learning and predictive coding. Hope you can join us.