Less than a decade ago, the Federal Rules of Civil Procedure (FRCP) turned litigation on its head by including amendments defining e-discovery. Now, as the amount of data to sift through explodes due to social media and other big data sources, legal teams are turning toward predictive coding to find relevant materials. However, experts caution that while predictive coding tools can speed the scanning process, they currently can't replace human review.
Predictive coding isn't a new technology, according to Tom Barnett, managing director and e-discovery practice leader at New York-based digital risk management and investigation firm Stroz Friedberg. "Predictive coding -- as useful as it is and can be -- is twentieth-century technology being used to solve a nineteenth-century problem, based on eighteenth-century mathematics," he said.
For the legal field, however, predictive coding is cutting-edge technology. And despite its age, predictive coding is very useful in the legal industry, Barnett said. Combined with other technologies such as data extraction, predictive coding can actually identify entities or pieces of data to find the proverbial needle in the haystack during discovery, he said.
The volume of data in large litigation matters such as product liability necessitates using predictive coding, but smaller bankruptcy cases have also found predictive coding to be advantageous, said Jenna Aira-Ventrella, managing director of e-discovery at New York-based BDO Consulting. "We're using and applying [predictive coding] in bankruptcies where the estate of the trustee doesn't have a huge amount of money to spend on e-discovery," she said.
Predictive coding helps find rogue documents
While reducing costs is one advantage of using predictive coding, another benefit is finding documents that a human reviewer may have missed, Aira-Ventrella said. The system needs to be "trained" to understand nuances in the documents, but once the software is programmed, it can find the same context in another document that has the same language and context nuances that the reviewer seeks, she said.
It's reliable in the sense that, whatever you tell it to do, it will do many times over.
president, SunBlock Systems
"This is key because you might get a higher level or more focused subset of comments that are germane to [one] particular issue, whereas a reviewer who is not as trained or a junior reviewer might have missed the intricacies or perplexities of the issues," Aira-Ventrella said.
Predictive coding also offers more consistency than human review if the system is properly programmed. One investigator may tag a document as nonresponsive, but another, more senior team member would categorize the document as relevant, Aira-Ventrella said. The system, if trained by a small core group of senior reviewers, is more likely to identify the documents most relevant to the litigation and minimize the identification of non relevant and false positive documents, she added.
The key to consistency, however, is making sure the right people are training the system, said David Sun, president of Washington, D.C.-based consultancy SunBlock Systems. "It's reliable in the sense that, whatever you tell it to do, it will do many times over," he said. For example, a lawyer may tell a predictive coding system to only look for documents containing keywords, while a programmer may expand the keyword searches to include misspellings.
Courts see the value of predictive coding
Despite some lawyers' reluctance to turn to technology, judges are seeing the value of using predictive coding when responding to document requests. The most significant has been in Global Aerospace Inc. et al., v. Landow Aviation, L.P. dba Dulles Jet Center, et al, Sun said. In that case, Judge James Chamblin of the 20th Judicial Circuit Court of Virginia ordered that the defendants could use predictive coding instead of human review to examine 2 million documents. "What that does is open the door for lawyers in this case and all other cases out there," he said.
That decision also allows attorneys to pass the buck to the judge if the predictive coding system doesn't find all relevant documents, according to Sun. "Lawyers deal with finding fault, and the last thing they want is to be declared at fault …. Now they can assign blame to the judge in this case," he said. "If the technology stinks, they can point to the judge, saying 'you approved it; sorry it didn't work out'."
More on e-discovery strategy
Social media data gains importance in e-discovery
Podcast: The keys to e-discovery strategy
Predictive coding court approval also creates opportunities for lawyers to agree how much they will use it and to what extent, particularly at Rule 26F conferences where both sides lay out their discovery plans, Sun said. The lawyers can determine what documents to search and at what level of relevancy, like 70% or 80%, they will consider as responsive to their discovery requests. "Lawyers will argue about a cutoff number. At the end of the day, it allows them, in a way that lawyers can work with, to argue over a number," Sun said.
For example, during the Rule 26F conference, the lawyers could compromise and decide that both sides should produce documents that are in PDF format and keyword-searchable, and they would use predictive coding to identify documents that are at least 80% relevant in response to discovery requests, Sun said.
"It's another thing that potentially you could agree upon early enough in advance so that you don't worry about bickering and fighting down the road," he added.
About the author:
Christine Parizo is a freelance writer specializing in business and technology. She focuses on feature articles for a variety of technology and business-focused publications, as well as case studies and white papers for business-to-business technology companies. Prior to launching her freelance career, Parizo was an Assistant News Editor for searchCRM.com.