Today's General Counsel, V15 N1, Spring 2018

Page 22

SPRING 2018 TODAY’S GENER AL COUNSEL

E-Discovery

Turning E-Discovery Concepts into Practice By Mike Hamilton

E

20

very year seems to usher in a new technology, a surprise in case law, or some other big idea that promises to revolutionize ediscovery. More often than not, the big concepts don’t fundamentally change the way e-discovery takes place. True revolutions are not common, but a lot of real value can be lost if we flatly dismiss “revolutionary” ideas. Often practitioners just lack practical tips to gain incremental benefits in efficiency or outcomes. So what are some concepts that have produced a lot of buzz — if not revolutions — in e-discovery and how might professionals find real value in them? AI AND PREDICTIVE CODING

Whether you talk about predictive coding, machine learning, AI or technology assisted review (TAR), few would dispute that the use of computers to review and classify documents during e-discovery is here to stay. The question is, by not

technology have had a negative impact on the benefits it can provide. In a recent edTalk sponsored by Exterro, Inc. and Georgetown Law CLE, Robert Keeling, Esq., Co-Chair of the E-Discovery Group at Sidley Austin, offers advice to legal departments and law firms that aren’t getting the results they expect from predictive coding: Be selective during document collection. Too many users, Keeling explains, “just throw everything into the so-called predictive coding hopper. In other words, they think, ‘I’m using predictive coding, so I don’t need to exercise any kind of discretion… because the model is magic and it’s going to figure it all out.’ ” Overcollection makes it more difficult for algorithms to recognize responsive documents. Use multiple models or techniques. Often one model will suffice to assess multiple kinds of data. But not always. Language differences, e.g., terminology differences between the United States

Tactics like seeding the sample with responsive documents can help train the algorithm. distinguishing between these concepts, has the legal profession failed to take the best advantage of what they offer? Given a green light by the courts in a series of decisions from Da Silva Moore and Rio Tinto through the recent Winfield v. City of New York, the federal bench contains some of predictive coding’s most prominent advocates. But in practice, it doesn’t always deliver the results its proponents hope for. Popular conceptions of AI as a near-magical

and the United Kingdom or Australia, might necessitate multiple review models. The same lesson holds for sampling. “Think about other sampling techniques,” Keeling advises, “for example, clustering first and then sampling across different clusters and making sure that at least part of your sampling captures all clusters.” Customize your workflows. There may be situations — for example, an internal investigation — when users can

defensibly sacrifice recall for precision. “That may not be defensible if you’re litigating a civil matter, but it may identify all the documents that are important in an internal investigation while saving significant expense,” Keeling says. Predictive coding has use beyond responsiveness reviews: quality control, privilege reviews, and even prioritizing attorney review. Be flexible. Adjusted parameters, algorithms and even sample sets, especially in low-richness environments, can yield better results. Tactics like seeding the sample with responsive documents can help train the algorithm. Experiment with default settings in a test environment to gain insight into how the algorithm works and what adjustments in parameters can provide. As Keeling explains, “The predictive coding process involves an attorneyidentified document that trains an algorithm to create a model that identifies and classifies the rest of the documents in the data set.” If the technology fails to deliver results, always consider user error among the reasons why. COLLABORATION IN LEGAL PROJECT MANAGEMENT

In the world of the law, collaboration may seem like a fantasy. In court, there are winners and losers. Inside firms, attorneys compete for partnerships. Even law school has a reputation for being exceptionally competitive, as students at the top of the class vie for prestigious clerkships. In fact, however, collaboration can and probably should exist in the legal profession, at least more often than it does. Heidi Gardner, Ph.D., Lecturer on Law at Harvard Law School, recently authored a book on the topic, Smart Collaboration: How Professionals and Their Firms Succeed by Breaking


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.