I often wonder whether the Googleification of legal research isn’t a terrible thing for the profession (at least in this stage of the technology’s development). In law school, I was a master of Boolean searching. I thought about my research question, figured out which words probably appeared closest to other words, and crafted a narrow and specific search.
Somehow, when I became an appellate attorney and had access to WestlawNext through my firm, all of that training went out the window. I think it seemed easier to just type what I wanted into a single search box and hope it returned a useful case (motion to start calling this the “click and pray?”). I got into the habit of assuming the algorithm was better than I was at crafting a search, but the truth is that right now, they’re not. Consider this: if a natural language search was as effective as a keyword search, it would be superfluous to pay an attorney for legal research. Most issues requiring us to do research implicate our ability to access information we’ve stored on relevant subject areas either in law school or in practice. Good algorithms can mimic this behavior to some degree — if you type “motion in limine procedure” into a natural language search on a legal platform, it’s conceivable that a such an algorithm might recognize that either based on your previous searches, or based on the aggregated knowledge of all searches, you’re probably looking for criminal cases (even though you didn’t mention it). Then it might recognize that you’re looking for some instructions on how to file such a motion and suggest some synonyms for the search. If this is my starting point for research though, I’m putting a lot of stock into an algorithm I can’t see (aside: I’d love to show our users the search query produced from a natural language search — I’m talking to our developers about making that a reality). I can sort of bypass most of this uncertainty by playing with a search like: defend! (“motion in limine” or MIL) /10 (procedure! or protocol! or formula! or method!) not methodolog!
That search looks complex but it’s probably going to give me results a lot closer to what I’m actually looking for than any natural language search is going to return.
I’m not at all saying we shouldn’t be using and improving natural language searching. In a pinch it can be useful. What I am saying is that as it stands right now natural language searching is not a replacement for a well-crafted Boolean search. And that’s a disconcerting revelation because out of my previous three law clerks, not a single one was well-versed in Boolean searching. I don’t know whether natural language searching has become the de facto LRAW teaching method or whether my sample is just skewed, but it’s worrisome. I would pay a lot of money for the ability to use a proximity operator in a Google search, but at this time they don’t offer that feature. Why, then, should we be taking a step backward? Yes, the ultimate dream solution would be telling a Watson-like computer the facts of our case, having it issue-spot, and then suggest a relevant search, but at the moment such a solution doesn’t exist. The problem isn’t so much that it doesn’t exist, but that I used to assume that it did. I don’t do that anymore and the time I spend researching has declined fairly dramatically.
Incidentally, I teach an introduction to Boolean searching webinar for Fastcase about once a month (to start again in January). If you’re interested in picking up the basics, I highly recommend you go ahead and like our Facebook page so that you are the first to know when we announce the dates for the 2014 year.
Happy /3 research!