The first thing we do, let’s kill all the lawyers.
Kidding! Let’s not kill all of them. Some of my best friends are lawyers, and they can be extremely handy. As a journalist, I would prefer to keep some around for interviews -- they are so precise and eloquent.
Even so, new technological developments make it seem likely that there are far too many of them. Innovations in the way the legal industry processes and sorts data could put thousands of low-level lawyers out of a job in the next few years.
It all goes back to a process called “discovery,” which is a defining trait of the American legal system. In the US, a party in a civil lawsuit can demand that its opponent share relevant e-mails, answer written questions under oath, even undergo a medical examination before the trial. This pre-trial discovery process aims to encourage the disclosure of all information relevant to the case, in order to increase the chance that the truth will come out in court.
In the days of yore (before the internet, pictured below), this information took the form of letters, memos, and other pieces of paper packed into dozens of filing boxes. Some young, lowly associate would have to sort through the documents one by one, decide if they were relevant to the case, and share them with the opposing counsel.
This wasn’t an easy process to begin with, but it became much more complicated with the digital age. The internet, personal computers, cloud computing, and other technologies have caused the number of digital records to skyrocket, with no signs of slowing down. The amount of data in the world now roughly doubles every two years, and by 2020 there will be nearly as many digital bits in existence as there are stars in the universe.
Yet – and this is the crazy part – the way lawyers sort through all this information remains pretty low-tech. Firms generally use some software to filter documents, but most of the work is still done by humans. To keep costs under control, firms now export the reviewing process to “contract reviewers,” armies of otherwise-unemployed lawyers who click through 50-100 documents per hour for up to 12 hours a day.
These attorneys -- many of whom are recent graduates of third-tier law schools, often with $100,000 or more in loan debt -- may earn as little as $8/hour, without benefits and overtime pay. They also frequently find themselves out of a job, as projects only last a few weeks or months. Yet the steadily rising number of law school graduates and still-sluggish economy ensure a steady supply of highly educated meat popsicles to do these jobs.
In cases where information security, attorney-client privilege, and logistics aren’t an obstacle, some firms are outsourcing the work, turning to English-speaking lawyers in countries like India and the Philippines. Junior attorneys in India can make about $8,200 per year – a rate not even South Carolina can beat.
Even with these armies of low-paid humans, however, the cost of reviewing documents continues to spiral upward along with the amount of data to be reviewed. Document review accounts for over 70% of the total costs of discovery, and some claim that the process is now so expensive that it prevents parties from litigating legitimate disputes.
Can’t we come up with something better?
Clearly, the legal industry is ripe for technological disruption. And it is on the cusp of it, thanks to something called predictive coding. Computer programs are increasingly able to “learn” desired traits and organize documents based on them, slashing the costs of manual review.
Here’s how that works. First a computer draws a statistically representative sample of a few hundred or thousand documents from the overall data set. An experienced lawyer examines this “seed set” and decides which documents are relevant. The program then tries to replicate the lawyer’s choices, assigning scores to another set of documents based on the probability that a document matches the desired characteristics. The lawyer examines and critiques its choices, helping to refine the program. The process continues until the software’s failure rate is reduced to less than 5% -- a rate that is far lower than a human reviewer.
Predictive coding has clear advantages. First, the cost is substantially lower. Obviously, people still play a crucial role in this process: A highly skilled and trained lawyer must work with the computer program to “teach” it what constitutes a useful document. But electronic discovery vendors report that predictive coding can reduce the documents to be reviewed by 50-75 percent, and the overall costs by 70-85%.
In addition, computer programs are actually far more consistent than their human counterparts. Reviewers tend to apply inconsistent standards from document to document, and the problem is multiplied when a team of reviewers is used. Computers are also, obviously, much faster.
So why hasn’t predictive coding been adopted already? It is better suited for some situations than others: It's more useful in cases involving large troves of documents, and it can only be used to review text, not pictures, databases, or spreadsheets. Another barrier is the lack of judicial decisions on predictive coding: A handful of courts have ruled that the use of predictive coding is acceptable, but more judicial decisions are needed before the technology is widely adopted. Finally, the legal industry seems to be moving slowly to adopt the technology, given that it will lose a substantial revenue stream.
The technology has big implications for improving the efficiency and lowering the cost of litigation. The sad part of the story, of course, is the number of already-low-paid lawyers who will soon find themselves without even the menial work of contract review. This change will only intensify the current oversupply of lawyers in the US. It's hard to say exactly how many Americans will be affected, but the figure is at least in the thousands.
Are you a law school graduate who works as a contract reviewer? I'd love to talk with you. Please get in touch via the comments section below or Twitter.
