I am an unabashed technology fan, but I’m also a lawyer and, therefore, a professional skeptic. So, when I read speculation about the one of the artificial intelligence (AI) packages taking over the work of law firm junior associates in the near future, the geek in me is intrigued and the skeptic in me says “no way.” Right now and for the foreseeable future, the skeptic is going to win. In fact, right now the chances of getting a computer to do first year associate legal research in the near future are about the same as getting Shakespeare verses from a monkey with a typewriter.
Let’s start with what is possible in the real world today. Siri, Contara, Google Now and Watson each can do wonderful things when it comes to tracking down information in data sets. They have, in varying degrees, the ability to take a query, parse the query, search databases, and come back with information relevant to the query. In some cases, they just deliver web sites that seem to have relevant information. In other cases, they can go much further. Some of the sophistication in these packages lies in their abilities to understand queries with some ambiguity built into them. The more ambiguous the query, the more the software must work to decipher the query, and that is much more difficult than simply searching databases.
The challenge becomes even greater when the complexity of the query grows. In other words, as we move from data searching to compound analytics, the software can’t simply go to one database and extract the answer. It may have to tie together multiple databases and knit an answer. Parsing the query also becomes more difficult, since compound queries typically include more ambiguity.
If you read Wired’s story about Viv, an AI package in development that, from the sounds of it, would blow the existing contenders out of the water, you will find a nice infographic showing how Viv would work through and answer the query, “On the way to my brother’s house, I need to pick up some cheap wine that would go with lasagna.” Using database searches and matching the results, Viv finds the route to “my brother’s” house, locate wine shops along the way, picks red wines that go with lasagna, and finds the prices of the wines.
AI and the First Year Associate
Ron Friedman and Alan Rothman discuss the logical question of interest to lawyers: could Viv take over doing basic legal research? Basically, could Viv become a first year associate? The same question has been asked about other AI efforts, but since Viv aims to play in a different league, I’ll focus on what I think would be the barrier to Viv taking over that first year associate’s job.
Legal research is quite complicated, but for purposes of this post I’m going to grossly oversimplify it by dividing it into two very large buckets. The first bucket is the easier of the two. It involves queries such as: find a case that supports the statement I am making in my brief. If I want to say “granting summary judgment in a patent infringement case is appropriate in the Federal Third Circuit when …” I need a case supporting that statement. As I understand Viv, she would parse the sentence into something like “what is summary judgment,” “what is a patent infringement case,” “what is the Federal Third Circuit,” and “what is appropriate.” She would then do something similar to what any of us would do, and follow the databases to find that case in the Third Circuit that says, “granting summary judgment is appropriate in a patent infringement case when the following test is met …”
Of course, things become more complicated if there isn’t such a case in the Third Circuit. Now, Viv must reason that the general standard for summary judgment in the Third Circuit also should apply to a patent infringement case. To do that, Viv would have to know whether a patent infringement case is distinguishable from other cases in a way that is meaningful for determining summary judgment. Even this simple question can quickly become confusing.
Let’s go back for a moment to the “cheap wine for a lasagna dinner at my brother’s” example from the Wired article. Viv recommended wine stores selling cabernet’s ((Viv’s wine suggestion) on the route from the requestor’s house to her brother’s house and found the prices for those wines. While Viv’s ability is amazing, it misses the human elements. Would a different wine choice be more inspired? Was the requestor’s brother serving tomato sauce lasagna or white lasagna? Should the requestor also bring a white wine, because her brother’s girlfriend is allergic to red wine?
You could argue I’m just making the initial query more complicated by adding facts. But, that is what humans do, although often subconsciously. We simplify what we are doing when reporting it out loud, but when we actually perform the task we take into account those unspoken factors. Con=mbining database research on discrete factors into an answer isn’t the same as legal research.
Similarly, when lawyers do research they don’t articulate all of the factors they are using when they select cases, they rely on factors not articulated. It is the factors not articulated, but used in the research, that can distinguish the quality of product returned. Unless the person asking Viv to do the legal research is very detailed when giving Viv the inquiry, and unless Viv can handle extremely complicated inquiries for even the most basic case support, I think it is unlikely Viv will be doing legal research anytime soon. I’m also skipping over the fact that frequently legal research involves finding the analogous case, not the case directly on point (because no such case exists), and that type of reasoning is well beyond Viv’s capabilities.
Respect, But Don’t Fear
If Viv and other AI software won’t be doing legal research soon, should we as lawyers care (er, worry) about these packages? I think the answer is yes, and for several reasons. First, Viv may not get to legal research, but undoubtedly a Viv in the not-too-distant future will get there. The discussion about computers doing some form of research is not one of “if” but of “when.” Second, the path from Viv to that future researcher probably won’t be linear. The path will involve leaps and plateaus, but we should be prepared for the leaps because each one will be a bit more of technology encroaching on what lawyers formerly did. Third, even though Viv can’t do research there are things a Viv-like AI package soon will be able to do and even those things will impact the way we practice law.
Lawyers today already face a deluge of information. Software packages today are good at and quickly growing better at pre-sifting information. This isn’t legal research; it is sorting through hundreds, thousands or millions of documents to find those relevant to the inquiry. Obviously, eDiscovery is the leading area where we see this happening, but these skills are now being used in other (though related) areas such as due diligence. Vendors are starting to find ways to gather and structure the unstructured information sitting in corporate and law firm databases. As more data becomes readily accessible and “meaningful,” software will use it more effectively.
Right now, we all seem focused on when certain changes will occur. Instead of focusing on the unknowable, we should focus on what we can control. Eliminating non-value added activities in what we do frees up our time to focus on value added activities. The more we focus on value rather than effort, the more we have a chance of staying ahead of the technology curve.
[N.B. – Points for those of you who figured out the “monkey with a typewriter” reference is to the picture which shows one of the members of the 1960’s band Monkees sitting at a typewriter. Extra points if you know which member of the band is shown, and points deducted if you have to look it up on Google.]