Take those robot lawyer announcements with a pinch of salt. Mainstream media can be a bit trigger-happy to get that clickable headline out there. I have seen dozens of applications that have been named robot lawyers and there’s nothing wrong with that necessarily; as I have been a bootstrapped entrepreneur myself once, I really don’t mind some marketing gimmicks. But is it just creative marketing, or could we already have one or two real robot lawyers out there?
The answer to the question whether we already have a robot lawyer, of course, depends on the definition. Basic legal research can already be aided by lawbots that can shuffle through the whole body of case law faster than is humanly possible. Basic AI techniques have also been used in contract review and electronic discovery over a decade now. Machines powered by AI techniques and some algorithms can do this a lot faster and with a lower error rate than a human lawyer or a team of lawyers ever could. Some junior lawyers have already lost their jobs because of this development and many other freshly graduated lawyers won’t be hired because of it.
However, is that enough to announce that the robot lawyers are already out there? As a lawyer who has taken his time doing his two law degrees and have had time to ponder the true essence of one’s calling, let me try to describe what the robo-lawyer is trying to automate here and whether we could call it “lawyering”. Let’s start by describing what kind of cognitive processing goes through the mind of a lawyer.
For example, in the U.S. common law system, lawyers interpret the logic of older cases and reapply the logic to a new case. In continental Europe, where we deal more with statutes and codes, you might concentrate more on interpreting the logic and intention of the legislator. In both instances, we use logical syllogisms, analogues, and often a combination of a few different inductive and deductive methods to find and apply the law to the case at hand. All in all, it’s quite a mess. There are so many moving parts and it’s less systematic than you might think. Legal rules can be extremely vague and contextual, which complicates things exponentially. Statutes and codifications of law are not semantically machine readable in a sense that would automatically lead any current legal AI to the right conclusion.
As a side note, often judges simply decide on what is “fair” and “reasonable” to them and find the right legal reasoning to go with it afterward. The thinking process is suddenly reverse. This teleological process, where the reasoning follows the conclusion, is an interesting phenomenon already in itself. Consequently, there might not be a template for the AI to follow. Judges can just fill out the holes in their logic with another type of logic to make the argument aesthetically pleasing. The legal aesthetics aside, jurists often use more than mechanical logic. They can use their whole life experience and outside circumstances to come to their conclusion. That’s the legal realism of it, but I’m not even sure how many judges would readily admit that. All in all, law can be systematic, but only to a degree.
Some people argue that robot lawyers are not “robots” or “lawyers” by definition, which in a strict sense of the terms is true (as there is neither physical body of a robot nor the required formal credentials of a lawyer). But to define “robot lawyer” as what the general public could call a robo-lawyer is not necessarily bound by those definitions. I would imagine that people simply expect a certain amount of autonomy from the robot, not being under much more supervision than a junior lawyer in a law firm would be.
At the moment, there is no artificial general legal intelligence (let’s coin it as AGIL) that could do all the lawyering that an average lawyer does, not even close. However, that criteria may be unreasonably harsh. We could also just judge the robot by the results it produces and compare that to a work product of a human lawyer. Even so, to my knowledge, we don’t have anything comparable to that in any field of law. My prediction is that most law students can feel safe for another 10 years.
Why aren’t they here yet?
If you think about it, the world is already full of robots building everything from cars and mixing perfect cocktails (video) to 3D printing houses (video). Robots are also active traders in the financial markets – accounting most of the trading volume, 75 % in fact – and even acting as reindeers (courtesy of Boston Dynamics). In my mind, you could easily call them robots. They are all legitimate robots and they do their work well, better than humans at least. However, they are still just extremely specialized artificial intelligence and pretty basic as such. It’s not quite enough to compile a robot lawyer from these same parts.
Think about how we depicted robots in the movies in the 1960s. The robots and bots of today are something completely different. However, that is where I think we are with “robot lawyers” at the moment. We are still figuring out what a “robot lawyer” should look and feel like. Perhaps a chatbot or possibly something else? At the moment, we can already automate variety of tasks with AI, but there are a couple of technological obstacles we have to work around before we can have a clear vision what is truly possible for robot lawyering and what is not.
The first of the two most restricting hurdles in building a robot lawyer is an AI technique called Natural Language Processing (NLP). Processing natural language, and even more so formal legal language, is not quite there yet. Often times NLP can’t yet understand what “it” or “them” in the sentence refers to. It’s a handicap which wouldn’t allow a lawyer to pass the bar exam (or these last two sentences). You intuitively know what that italicized “it” at the beginning of the previous sentence refers to but how would you program that in for the AI to get it consistently right?
The other aspect is that lawbots have to go to a law school as well. Legal AI simply applies its past experiences and training history to new situations and determines whether a result A, B, C or N is more likely. Without rigorous training, without a set of thousands of questions and answers, the robot lawyer can’t reach the lowest of bars in processing natural language. Moreover, legal language with its intricate way of defining every term and concept is even trickier. You could easily write a book defining what a certain term means in a legal context. This is something that a “robot lawyer” would have to understand explicitly or intuitively; human lawyers accompanied by their real life experiences usually have it both ways.
Long story short, in July 2017, we don’t yet have a robot lawyer. We probably won’t have one for a while. Moreover, my personal guess is that (alongside centaur pairings of human lawyers and lawbot assistants) a semi-autonomous robot lawyer of the future will be a collection of different systems and specialized lawbots working together – not too dissimilarly how law firms work at the moment.
Or it could be that my imagination just doesn’t reach further than that. Indeed, it is getting harder and harder to predict technological advancements ten or twenty years from now. For example, in the year 1300, it was a lot easier to predict what the advancements in technology might be by 1320 (perhaps a pointier quill pen). It was even easier in legal technology (i.e. nothing happened in the next 100+ years before the printing press). And then you can compare it to the year 1990: Few people could have predicted where we were in 2010 and now it’s practically impossible to predict where we will be in 2037. That is just indicative of the exponential nature of technological advancement, as one technology enables two or three others (and so on).
As a side note, even if you don’t believe in some people – wiser than me – who talk about technological singularity where robots take over and your only defender in a court might indeed be a robot lawyer, one thing is for sure: Technological progress is not linear and that applies to legal technology as well. Keep your eyes open particularly for the development of NLP and self-teaching AI.
How do you spot a robot lawyer?
If you want to be there to witness the first robot lawyer, look for the legal fields and customer segments that appreciate full automation as a way to save in legal fees. For some, it’s not just saving money but it can easily be the difference between using legal services or not. That said, look for legal services for small businesses and startups, legal aid services, and automated legal services targeted for middle-income consumers. Even though not really a robot lawyer, for example, look for affordable services like separate.us that is able to help with divorce procedures.
As said, media can be a bit trigger happy to get that clickable headline out there, so there are already dozens of applications that have been named robot lawyers. And by some definition, you could even call them that, but what most people would consider a “robot”, or autonomous legal AI, that is not out there yet. If there were a robot lawyer, I would have bought one already. I would just need that $10 million loan first.
Even though robot lawyers are not here yet, the good news is that we will keep you updated at lawbots.info. As a little bit of a plug, we will get our first consumer lawbot awards out – yep, we avoid the term “robot lawyer” – later in early 2018 so look out for the announcements. We have already pointed out some great free and affordable services for startups to use. We’ll update the awards annually, so feel free to suggest a new category of automated legal services for startups or another kind of lawbot you have found useful.
I believe that a robot lawyer would also be the ultimate pinnacle of artificial intelligence. As we can easily argue that we already have robot doctors, journalists and investment advisors (feel free to correct me), probably one of the last things missing would be a robot lawyer. All in all, we are living exciting times in the legal practice and everyday lawyering. And that is not a thing you can often say outside your lawyer circles.
Thank you for reading the article. Also, I’m humbled that the web crawler bots and search engine AI (or Twitter’s new AI) judged this article worthy enough for you to find it. Feel free to share it. You can follow us or share the article on Twitter here. Also, despite the AI that proofread the article, human errors might occur.