February 6, 2010 - Dulles Jet Center
~
Thomas Gricks, right, of Schnader Harrison Segal & Lewis, and Karl Schieneman, an expert in e-discovery, during a presentation of predictive coding.
Photo Credit: Larry Roberts/Post-Gazette
By JOE PALAZZOLO
The Wall Street Journal
When the roofs of three jet hangars in Virginia collapsed under heavy snow and crushed 14 private jets in 2010, the owner of the hangars prepared for the inevitable lawsuits.
Landow Aviation preserved about 8,000 gigabytes—the equivalent of about eight new desktop computers filled to the brim—of company emails, documents and other electronic information about every aspect of its operations that it might be asked to produce.
After insurers and the aircraft owners started filing lawsuits against Landow during mid-2010, the company identified a batch of about two million electronic documents it would need to sift through for evidence of its possible liability in the roof collapses.
Rather than hire dozens of lawyers to read the documents, the company asked a judge to allow a computer program to do much of the initial work. In late April this year, Loudon County Circuit Judge James H. Chamblin granted permission for Landow to use "predictive coding," a general term that refers to computer programs that use algorithms to determine whether documents are relevant to a case.
The mounting costs of high-stakes litigation are largely driven by so-called discovery, in which parties exchange documents before trial that they deem relevant to the legal claims. In the typical discovery process in big litigation, an army of lawyers reviews each document before turning it over to the other side. Industry experts peg the price of this review at more than $1 per document. Supporters say predictive coding is as effective as a set of human eyes—and much cheaper.
The programs essentially work like this: After documents are loaded into the program, a lawyer manually reviews a batch to train the program how to recognize what is relevant to a case. The manual review is repeated until the program has developed a model that can accurately predict relevance in the rest of the documents.
Thomas C. Gricks III, a Pittsburgh-based lawyer representing Landow, estimates that his client would pay about one-tenth of what it would have paid if lawyers reviewed all two million documents.
Landow's lawyers will still review the documents that the computer program flags as relevant, which in a typical case is about 10%, Mr. Gricks says.
"One way or the other, you're going to have lawyers that are going to look at that 10%," he says. "The question is, how do you generate that 10%? Do you use lawyers or do you use a tool that's much more cost effective?"
It is unclear how widespread such technologies have become in litigation, not least because parties aren't obligated to tell each other when they used such computer programs to assist them in discovery.
Predictive coding is one of several computer-driven tools that lawyers use. The most common is keyword searching, in which documents are loaded into a program and lawyers input search terms to find relevant documents. Keyword searching is used alongside human review.
But some consultants say that more corporate clients have started to warm to the idea of letting machines do more of the heavy lifting in litigation, after a landmark ruling in a discrimination lawsuit against French advertising company Publicis Groupe SA and its public-relations unit in February.
Predictive coding should "be seriously considered in large-data-volume cases where it may save the producing party (or both parties) significant amounts of legal fees in document review," Magistrate Judge Andrew J. Peck in Manhattan said in his ruling in the case.
Greg McPolin, an executive at the legal outsourcing firm Pangea3, which is owned by Thomson Reuters Corp., TRI -0.07% says about one-third of the company's clients are considering using predictive coding in their matters.
"We would have our heads buried in the sand if we did not embrace this technology," he says.
As companies and their employees create more data, the haystack is growing geometrically. The world's information is more than doubling every two years, with an estimated 1.8 zettabytes created in 2011—the equivalent of about 1.8 billion desktop computers—according to the data-storage company EMC Corp.
As recently as a decade ago, corporate-law firms conducted large-scale document review in-house, assigning young lawyers to the task who billed several hundred dollars an hour. But in recent years legal staffing companies have proliferated, supplying law firms and legal departments with temporary lawyers who are paid as little as $25 to $30 an hour to review documents.
Predictive coding isn't expected to replace this new breed of low-rent lawyer, but it could significantly reduce their numbers during one of the worst employment markets in nearly 20 years, industry experts say.
Several studies have shown that predictive coding outperforms human reviewers, though by how much is unclear. A widely cited 2011 article in the Richmond Journal of Law and Technology analyzed research on document review and found that humans unearthed an average of about 60% of relevant documents, while predictive coding identified an average of 77%.
The research also showed that predictive coding was more precise, flagging fewer irrelevant documents than humans did.
"Human readers get tired and make mistakes. They get fatigued," says David Breau, an associate at law firm Sidley Austin LLP who has written about predictive coding.
"Computers excel at sifting through a big pile of stuff and sorting it into categories," he adds. But he believes that lawyers are still needed to review the documents once they are sorted before turning them over to the other side.
Paul Neale, chief executive of DOAR Litigation Consulting, for one, doesn't. Mr. Neale's DOAR Consulting is advising Publicis Groupe and its public-relations subsidiary in its discrimination litigation.
"If you're going make the argument that it's more accurate than humans and that you should rely on it to reduce costs and time, then you should be relying on it fully," he says. "You can't get half pregnant."
Others are less convinced. "No computer program is an adequate substitute for having human beings review and sort the documents," said the group of lawyers representing the companies suing Landow, in an April court brief.
The lawyers either declined to comment or didn't respond to requests for comment.
Brandon Daniels, president of managed services at Clutch Group, another legal outsourcing firm, says the technology will have to prove itself in court.
If a company using predictive coding fails to turn over a crucial document in a case, for instance, "that could put this entire leap forward in jeopardy," he says.
"Everybody would just blame the predictive coding rather than the sophistication of the lawyers," he adds.
Source: http://online.wsj.com
When the roofs of three jet hangars in Virginia collapsed under heavy snow and crushed 14 private jets in 2010, the owner of the hangars prepared for the inevitable lawsuits.
Landow Aviation preserved about 8,000 gigabytes—the equivalent of about eight new desktop computers filled to the brim—of company emails, documents and other electronic information about every aspect of its operations that it might be asked to produce.
After insurers and the aircraft owners started filing lawsuits against Landow during mid-2010, the company identified a batch of about two million electronic documents it would need to sift through for evidence of its possible liability in the roof collapses.
Rather than hire dozens of lawyers to read the documents, the company asked a judge to allow a computer program to do much of the initial work. In late April this year, Loudon County Circuit Judge James H. Chamblin granted permission for Landow to use "predictive coding," a general term that refers to computer programs that use algorithms to determine whether documents are relevant to a case.
The mounting costs of high-stakes litigation are largely driven by so-called discovery, in which parties exchange documents before trial that they deem relevant to the legal claims. In the typical discovery process in big litigation, an army of lawyers reviews each document before turning it over to the other side. Industry experts peg the price of this review at more than $1 per document. Supporters say predictive coding is as effective as a set of human eyes—and much cheaper.
The programs essentially work like this: After documents are loaded into the program, a lawyer manually reviews a batch to train the program how to recognize what is relevant to a case. The manual review is repeated until the program has developed a model that can accurately predict relevance in the rest of the documents.
Thomas C. Gricks III, a Pittsburgh-based lawyer representing Landow, estimates that his client would pay about one-tenth of what it would have paid if lawyers reviewed all two million documents.
Landow's lawyers will still review the documents that the computer program flags as relevant, which in a typical case is about 10%, Mr. Gricks says.
"One way or the other, you're going to have lawyers that are going to look at that 10%," he says. "The question is, how do you generate that 10%? Do you use lawyers or do you use a tool that's much more cost effective?"
It is unclear how widespread such technologies have become in litigation, not least because parties aren't obligated to tell each other when they used such computer programs to assist them in discovery.
Predictive coding is one of several computer-driven tools that lawyers use. The most common is keyword searching, in which documents are loaded into a program and lawyers input search terms to find relevant documents. Keyword searching is used alongside human review.
But some consultants say that more corporate clients have started to warm to the idea of letting machines do more of the heavy lifting in litigation, after a landmark ruling in a discrimination lawsuit against French advertising company Publicis Groupe SA and its public-relations unit in February.
Predictive coding should "be seriously considered in large-data-volume cases where it may save the producing party (or both parties) significant amounts of legal fees in document review," Magistrate Judge Andrew J. Peck in Manhattan said in his ruling in the case.
Greg McPolin, an executive at the legal outsourcing firm Pangea3, which is owned by Thomson Reuters Corp., TRI -0.07% says about one-third of the company's clients are considering using predictive coding in their matters.
"We would have our heads buried in the sand if we did not embrace this technology," he says.
As companies and their employees create more data, the haystack is growing geometrically. The world's information is more than doubling every two years, with an estimated 1.8 zettabytes created in 2011—the equivalent of about 1.8 billion desktop computers—according to the data-storage company EMC Corp.
As recently as a decade ago, corporate-law firms conducted large-scale document review in-house, assigning young lawyers to the task who billed several hundred dollars an hour. But in recent years legal staffing companies have proliferated, supplying law firms and legal departments with temporary lawyers who are paid as little as $25 to $30 an hour to review documents.
Predictive coding isn't expected to replace this new breed of low-rent lawyer, but it could significantly reduce their numbers during one of the worst employment markets in nearly 20 years, industry experts say.
Several studies have shown that predictive coding outperforms human reviewers, though by how much is unclear. A widely cited 2011 article in the Richmond Journal of Law and Technology analyzed research on document review and found that humans unearthed an average of about 60% of relevant documents, while predictive coding identified an average of 77%.
The research also showed that predictive coding was more precise, flagging fewer irrelevant documents than humans did.
"Human readers get tired and make mistakes. They get fatigued," says David Breau, an associate at law firm Sidley Austin LLP who has written about predictive coding.
"Computers excel at sifting through a big pile of stuff and sorting it into categories," he adds. But he believes that lawyers are still needed to review the documents once they are sorted before turning them over to the other side.
Paul Neale, chief executive of DOAR Litigation Consulting, for one, doesn't. Mr. Neale's DOAR Consulting is advising Publicis Groupe and its public-relations subsidiary in its discrimination litigation.
"If you're going make the argument that it's more accurate than humans and that you should rely on it to reduce costs and time, then you should be relying on it fully," he says. "You can't get half pregnant."
Others are less convinced. "No computer program is an adequate substitute for having human beings review and sort the documents," said the group of lawyers representing the companies suing Landow, in an April court brief.
The lawyers either declined to comment or didn't respond to requests for comment.
Brandon Daniels, president of managed services at Clutch Group, another legal outsourcing firm, says the technology will have to prove itself in court.
If a company using predictive coding fails to turn over a crucial document in a case, for instance, "that could put this entire leap forward in jeopardy," he says.
"Everybody would just blame the predictive coding rather than the sophistication of the lawyers," he adds.
Source: http://online.wsj.com
Read more: http://www.post-gazette.com
