On April 25, 2012, District Judge Carter upheld United States Magistrate Judge Peck’s opinion in Da Silva Moore v Publicis Groupe et al. Justice Peck’s opinion, rendered earlier this year, made the headlines in the e-discovery bar when it became the first endorsement from a US federal judge of a litigation protocol for e-discovery that included predictive coding technology.
What is predictive coding?
In most jurisdictions across North America, parties to a dispute are required, by law or upon request, to identify and communicate all relevant documents in their possession in advance of the trial. This process can be expensive and time consuming.
Given that most of the documents generated these days are in electronic form, technology has been used to facilitate the process of identifying relevant documents. For example, keywords or Boolean searches are often used to identify a batch of potentially relevant documents, which is then manually reviewed by junior lawyers.
Over the years, a new tool has developed to help identify relevant documents: predictive coding. Predictive coding is a computer-assisted form of document review that allows more senior lawyers to first review a statistically relevant sample of documents for each issue and to attribute a relevance score to each document. These relevant documents are then fed to computer software which will use this sample (the “seed”) to identify other potentially relevant documents amongst the millions of documents that were not reviewed.
Samples of documents “predicted” to be either relevant or irrelevant by the computer software are then reviewed for quality control and a new seed is submitted to the software for a new round of electronic review. Several rounds (the discovery protocol approved by Justice Peck in Da Silva planned for a minimum of seven rounds) are usually made in order to properly “teach” the software to identify truly relevant documents.
Why is predictive coding considered more efficient?
Justice Peck explained that manual review, usually considered the gold standard in document review, provides at best results only as good as computer-assisted review and is much more expensive. In fact, Justice Peck cited several studies to the effect that manual reviewers underperformed software more often than not.
Justice Peck also explained that keyword searching was usually not as effective as predictive coding since, inter alia: (i) the requesting party needs to guess relevant keywords without necessarily being aware of the producing party’s own abbreviations and lingo for referring to potentially relevant concepts, people, parties, etc., and (ii) this process usually turns out a significant amount of irrelevant documents to be manually reviewed.
What should litigants learn from this opinion?
Justice Peck, who was extremely knowledgeable on these issues prior to rendering his opinion (to the point, in fact, that the plaintiff recently filed a motion for disqualification of Justice Peck on the basis that, among other things, he has spoken at legal technology conferences sponsored by defendant’s e-discovery vendor), concluded his judgment with a section several pages long entitled “Further analysis and lessons for the future” where he insisted that computer-assisted review is not “a magic, Staples-Easy-Button, solution appropriate for all cases” but that, on the basis of his opinion, “Counsel no longer have to worry about being the 'first' or 'guinea pig' for judicial acceptance of computer-assisted review.”
This last statement, coming from a judge from one of the most influential and active federal district courts in the United States, is likely to have a significant impact in the United States and also outside US borders.
The cost of entrusting junior lawyers with the task of reviewing millions of electronic documents is indeed a concern for all litigants in any jurisdiction. Concerns about costs of the discovery process have even become, in some cases, a bar to going to trial even when on the merits a party has no reason to try to reach for a settlement at all cost.
While the use of predictive coding does not make document review a “cheap” process by any standard, the fact that a reputable court has allowed the use of a potentially cheaper and more efficient alternative to manual document review may certainly feel like a sign of relief for litigants with a potentially significant amount of electronically stored information to review.
The fact that efficient computer-assisted review tools are now accepted in courts means that we are perhaps not so far from a time where discovery will cease eclipsing the issues of the merits.
Download US judge allows use of computer software for assessing relevancy in document review process (87kb)