Discovery for major litigation once meant junior attorneys would spend thousands of hours doing the mind-numbingly repetitive work of sifting through documents, identifying the ones that had to be produced for the pending lawsuit.
For clients, it usually meant sky-high bills to pay all those junior attorneys to do all that tedious sorting.
But advancing technology is increasingly putting the squeeze on what firms can bill for discovery. First there was the use of keyword searching—computers quickly scanning millions of documents to look for key words that might alert attorneys that a document is relevant to a pending lawsuit—to aid the reviewers’ task. Now the computers have gone one step further with “predictive coding” that drastically reduces the need for human eyes entirely.
In predictive coding, humans review only a small sample of documents, called a “seed set,” separating the relevant from the irrelevant. Computers then analyze the rest, looking for similarities to the seed documents that predict whether a document is relevant. Humans evaluate a sample of the computers’ work, letting the computer refine its process. The process is repeated until the computer can analyze the documents with sufficient accuracy, greatly limiting the hours required from expensive human reviewers.
Predictive coding reviews documents far more quickly and cheaply than human reviewers can. Now recent studies suggest that computers—which review even the dullest documents without ever becoming bored, tired or distracted—also classify them more accurately than humans can, too.
In just a few years, a technique that was once viewed suspiciously by judges has quickly started gaining acceptance. South Carolina courts have yet to weigh in on the question, but since the first court decision mandating the use of predictive coding in 2012, it has been endorsed—or imposed—in cases in New York, Virginia, Louisiana and Delaware. A survey of attorneys at Fortune 1000 companies released in March by FTI Consulting showed that 57 percent believed predictive coding would be a mainstream tactic by 2015.
Jeffrey Hahn, an attorney who manages large document review projects for North Carolina-based Womble Carlyle, says the use of predictive coding is growing and cutting costs for clients by reducing the need for human eyeballs. But computers have their limitations, he said.
“While I think these tools will make the process more efficient, I don’t think they’re capable of totally eliminating the need for human review,” Hahn said. “You still need to select a seed set of documents with human attorneys and do quality control so the computer learns and gets better, and you need to validate the results.”
For example, Hahn said that predictive coding is less effective on very short documents, and he had yet to see a program that could adequately determine which documents are protected by attorney-client privilege. He also noted that companies are now able to retain vastly larger amounts of electronic data, which increases the need for both predictive coding and human reviewers. Still, he said the technology is having a major impact on how firms handle discovery.
“I view it as a sort of tool in our toolbox that we can apply to make ourselves more efficient and provide document review in an even more cost efficient manner to our clients,” Hahn said. “Tools like predictive coding are helpful in continuing that trend to bring the cost of document review down.”
Given that discovery was once a steady revenue stream for big law firms, predictive coding promises to be a highly disruptive technology. The speed with which courts are accepting its use suggests that it may move rapidly from a cutting edge technology to a widely-used method to a required practice.
The downside, of course, is that while document review may have been crushingly tedious, it paid the light bill.
Follow David Donovan on Twitter @SCLWDonovan