From any perspective, 2016 was quite a year, but the eDiscovery world may remember this as the year Predictive Coding reached a tipping point. In US jurisprudence, this was one of the quietest years yet for TAR, with only 10 published orders mentioning Predictive Coding (the same volume as 2015), continuing a downward trend from the peak of litigation in 2013 (17).
However, for the first time ever, multiple international orders addressed Predictive Coding. What’s more, the content of these orders confirm that other jurisdictions are catching up with the general acceptance that TAR enjoys in the US.
The Major 2016 Predictive Coding Cases Abroad
The landmark Pyrrho Investments case was the first published order approving the use of predictive coding in the UK. Meanwhile, the Irish Court of Appeal affirmed IRBC, the first case approving the use of Predictive Coding in Ireland.
The Pyrrho holding was followed shortly after in Brown v. BCA Trading, where Mr. Registrar Jones remarked:
“I reach the conclusion based on cost that predictive coding must be the way forward.”
Before the year ended, Australia became the newest legal system to formally approve the use of Predictive Coding. First in Money Max v. QBE, the Federal Court of Australia (Victoria) approved the use of TAR and a protocol that laid out specific disclosure obligations. A month later, the Supreme Court of Victoria similarly approved the use of TAR and a party-driven protocol in McConnell Dowell Constructors v. Santam.
The Australian orders are relatively brief and straightforward (Money Max only has 3 paragraphs covering Predictive Coding). The core legal issue that they deal with is the general defensibility of TAR. The issues around protocols and disclosure are more nuanced and fact-specific; in both cases the orders indicate that the parties had come to court with a mutually agreed upon protocol.
Issues in US Litigation: Equilibrium on Cooperation
Back in the US, the judicial landscape remained quiet for the first half of the year. In Re: Bair Hugger—which features a remarkable protocol we will discuss at LegalTech 2017—issued in July. Within days, Dynamo Holdings 2 was published. This case featured a twist worthy of an M. Night Shyamalan: it turns out that even though receiving counsel had practically run the TAR process, they still weren’t satisfied.
Nonetheless, Judge Buch again approved the use of Predictive Coding and the specific process applied, and denied the IRS Commissioner’s motion to compel additional discovery. Shortly after, B&R Supermarket v. Visa also approved the use of TAR and laid out minimal disclosure obligations:
“[A]s early as reasonably practicable (and in any event prior to using such tools) it will disclose to the opposing parties the type of technology it will be using and a general description of the TAR methodology that will be used.”
The Novel Issue of 2016: Can You Compel the Use of TAR?
Once again, Judge Peck issued the most important Predictive Coding order of the year that addressed a particularly novel issue in Hyles v. New York. There, Judge Peck denied a motion to compel the defendant City to use TAR.
The order explained that while it is desirable to use TAR, “the Court cannot, and will not, force the City to do so.” Two months later, the Northern District of California, citing to Hyles and Judge Peck’s discussion, denied an identical motion to compel the use of TAR in In Re Viagra.
Is the US Past Peak Litigation?
In the chart below, I’ve plotted all the published US cases and orders that mention Predictive Coding since 2012. While this is, admittedly, a small dataset and an imperfect methodology (see below for details), it would be reasonable to theorize that litigants are becoming more comfortable with TAR and capable of settling disputes over discovery issues without judicial intervention.
This theory is backed up by the content of the published opinions (see Hyles or In Re Viagra); issues like defensibility simply aren’t litigated much anymore.
It’s important to note that a single case could include multiple orders that were counted towards the total. For instance, Da Silva Moore had 4 separate Predictive Coding orders, each one of which was counted towards the 2012 total. Furthermore, certain orders simply mentioned “predictive coding” in passing as a technology that the parties were to consider using.
Those were included as well. I reviewed each case manually (indeed, we’ve blogged about most of them here) as a quality check and identified only one false positive that was completely off-topic. I have compiled all my research in a case outline and a spreadsheet that I will be happy to share with you.