<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=513385&amp;fmt=gif">

By Greg Behan

Published on Wed, April 19, 2017

All posts by this person

Attorneys are still reluctant to utilize technology assisted review (TAR) in ediscovery even though numerous courts have approved this methodology. There is almost never a circumstance where a large group of contract attorneys will outperform one or two subject matter experts working with TAR in an appropriate feedback cycle. There seems to be a prevailing fear that leads to lack of adoption, and that hesitation most likely stems from the psychological acknowledgment that a computer can outperform its human counterparts. Lawyers, just like their historical ancillaries, are afraid of being replaced by machines.

Humanity has always had an interesting relationship with the machines created to replace labor. In early 19th century England, textile workers (the Luddites) organized and began smashing the machines that were automating weaving. As legend has it, around 1870, John Henry competed against the steam hammer in a contest drilling holes made for explosive charges that were blasting away rock to make a path for the Eastern American railway. He beat the machine, only to die later of physical exhaustion.  

In 1950, Claude Shannon (the father of Information Theory), set out to show that a brute force approach to teaching a computer how to play chess was impractical. The “Shannon Number” was the lower bound estimate of all the possible moves a game of chess could generate, and it was measured at 10120 power.  On May 11, 1997, in the 6th and final game of a chess match between World Chess Champion Gary Kasparov and IBM’s Deepblue, Shannon’s theory was tested. Kasparov had beaten Deepblue a year earlier, but the team at IBM had reprogrammed the machine by playing many of the world’s chess masters. Instead of using just brute force calculation, the supercomputer had studied human play through a course of machine learning to understand the patterns of the best players in the world. In the final game of the match, after only 19 moves, Kasparov realized the game was over and conceded. It was the first time in history a computer had beaten the world’s greatest chess player in a match. At the time, Deepblue had 256 processors and could calculate 200 million moves a second.    

In 2011, IBM’s supercomputer named Watson played the world’s greatest Jeopardy players on television. It was quite a spectacle watching this giant machine (Watson had 2800 processors and was about the size of 10 large refrigerators) compete against real people. The rules of the game prevented Watson from accessing the internet, so IBM’s team filled the memory with a myriad of information. Watson’s servers contained all of Wikipedia, the entire Worldbook encyclopedia, the whole Internet movie database, and every article ever written for the New York Times. It had an endless array of books, plays, magazines, and other public data sources from which to draw information. Even with all this processing power and memory, Watson still struggled with the puns, double meanings, contextual hints and other natural language nuances that make Jeopardy an entertaining game show to watch.

At first, Watson wasn’t very good at Jeopardy, but as it began to play against the human competition during its audition, it started learning. Every question and answer from every show ever broadcast was added to its memory. Watson started seeing patterns in the phrasing of questions and answers and then began to understand the larger semantic format of the game. When Watson finally played the best Jeopardy players ever (Ken Jennings and Brad Rutter) it wasn’t even close. Watson crushed them, earning $35,734 to Jennings’ $4800 and Rutter’s $10,400. 

Technology doesn’t replace humans as much as it enables us. Using TAR is the best possible way to retrieve information for litigation. And, just like the supercomputers in my examples, the machine is only as good as the lawyers who train it. The best lawyers today and in the future are the ones who work in conjunction with these machines. Tomorrow’s John Henry story will be the lawyer who beats the algorithm at trial. It will be the trial of the century…

For more information or to get in touch, please reach out to info@lhediscovery.com.

About the Author
Greg Behan

Hosted Solutions Project Manager

Greg is a trial lawyer with 10 years ediscovery and document review management experience. He focuses on automation and building processes that extract and present the narratives of litigation. He also has experience with forensic investigations, complex litigation, and technology assisted review.