by Erik Adams
Why hasn’t the artificial intelligence revolution come to current awareness in the legal industry?
When I started at my current law firm in 2002, current awareness was in a technological transition. Publishers of daily and weekly newsletters were starting to offer email versions in addition to print, almost as a novelty. We began to offer electronic delivery to attorneys as an alternative that offered a few advantages, chief among them not having to wait for a magazine while it sat in another attorney’s in box, because you had the misfortune to be lower down on the routing list. That, and being able to read the news on a Blackberry. 15 years later that transition is almost complete: we route very few print publications, and the number of email newsletters has exploded. If anything, we’ve gone too far in that direction, and I often hear from attorneys that they get too many emails.
The next big innovation came in the form of products like Manzama and Lexis Newsdesk, which use complex Boolean queries and keywords to generate custom made news feeds. Some attorneys love these services, but I know of at least one firm who has dedicated a librarian’s time to setting up these services for attorneys who don’t want to take the time to customize the news feed themselves. And the vast majority of attorneys prefer to have the news delivered via email, rather than going to the trouble of visiting a web site and reading in a browser. Blackberries have been replaced mostly been replaced with iPhones and Galaxies, but the attorneys still want the news in their hand, delivered from a variety of sources into the email app.
And there, technological innovation has stalled.
Which gets me back to my question: why isn’t AI being applied to the problem of news organization and distribution? We hear a lot about how Ravel Law, Lex Machina, and CARA have applied AI to difficult and time consuming legal research tasks in an effort to make them easy and quick. But the boring task of filtering a day’s news into something interesting and relevant has largely gone untreated.
Facebook recently announced that it is using AI to combat the problem of fake news, developing a system that will automatically flag articles with misleading or inaccurate information. And spam filters have employed Baysian style learning for decades. It seems to me that information overload is a similar problem: there are articles that should be promoted, and articles that should be withheld. If we’re not going to employ armies of librarians to do this job (and we certainly aren’t doing that at my firm), then perhaps someone can create an army of bots to do it for us.
I know enough about artificial intelligence to know that smart systems have to be trained, so I offer a simple plan to train a system. First, start with the articles from a well known, intelligently edited legal newsletter – for example, BNA’s Bankruptcy Law Reporter. Use this as the seed for “good” information. If a large sample of bad, non-bankruptcy news is needed, use any of BNA’s other newsletters. Any of the major vendors could do this. Why haven’t they, yet?