(Added: There are some excellent comments in the thread, well worth reading.) Walter Russell Mead and Tyler Cowen each point to this article in New York Magazine, describing a decision by US Magistrate Andrew J. Peck (SDNY) to allow the “use of predictive coding, a computer-assisted document review that turns much of the legal grunt work currently done by underemployed attorneys over to the machines.” The work performed by armies of temporary “contract lawyers”?
The task of combing through mountains of emails, spreadsheets, memos and other records in the discovery process currently falls on a legion of “contract attorneys” who jump from one project to another, employed by companies like Epiq Systems. Many are recent grads who are unable to find full-time employment, or lawyers laid off during the recent recession.
Scan. Point. Click. Repeat. That’s the job. Contract attorneys are paid by the hour to sit in front of a computer and review a mind-numbing sequence of uploaded documents. There are cramped, sunless rooms in law firms throughout the city, with rows of computers piled one on top of the other, and constant uncertainty as to how long each particular stretch of employment will last.
Predictive coding promises to make this job much more efficient over time – and drastically reduce the amount of work and number of contract attorneys employed.
Using the technology, a senior attorney familiar with the intricacies of a specific case reviews and codes a “seed set” of documents. An algorithm then identifies properties among the manually reviewed documents to code and sort everything else. Each document is assigned a score to indicate the likelihood it’s correctly coded.
Proponents say predictive coding is not only more accurate than using human reviewers, but also more efficient .... There’s no escaping the fact that as predictive coding is used more widely, the technology will reduce the overall number of documents to be reviewed and the attorneys needed to review them. Judge Peck noted the technology will require human review of less than 2 percent of all documents in an average case. His stamp of approval means that the document reviewer ranks may be culled sooner rather than later.
Mead adds a comment about what this means for the future of lawyering as a profession and as a safety ticket to the upper middle class. I would add the additional comment that survival lies in getting on the right side of algorithm production. That doesn’t necessarily mean that one has to write algorithms. We live in a highly complex economy with many niches. But whatever one’s niche, it helps a lot to be aligned on the side of the angels of creative destruction, rather than sitting in their path.
Software is getting smarter, and computers continue to grow more powerful. What we see now is only the beginning of a process by which the routine elements of legal work — and frankly speaking, that is where the bulk of the jobs have always been — can and will be automated. Not all young lawyers will be doomed. There will be some smart, entrepeneurial kids who figure out how all this computing power can allow a small, lightly capitalized firm to deliver high quality services at a breathtakingly low cost to selected clients. Those kids will do well.
Others will benefit from greater demand. Legal services are likely to get cheaper: there is a lawyer glut that is likely to grow, and the increasing capabilities of computers in the legal field mean that the amount of available legal brainpower will explode. Cheaper legal services mean that more people and firms will use the legal system and legal expertise in various ways: the lawyers and firms who figure out how to ride this wave will also do well.
Brilliant and creative lawyers will continue to do well. So will the marketers, the deal makers and the connectors. But law isn’t going to be the kind of safety play ticket to the upper middle class that it used to be.