Why Lawyers Won’t Be Entirely Replaced by Robots

Professor Michael Livermore’s work in computational legal scholarship may sound arcane to some, but legal practitioners should take note: Work like his will define the extent to which robots can replace lawyers.

Fortunately for those in the legal field, the University of Virginia School of Law scholar doesn’t see much chance of robots taking over the courtroom anytime soon, and his upcoming keynote address Thursday in Abu Dhabi — to natural language processing experts working with legal texts — will explain why.

“The process of making law is a human process,” Livermore said in an interview before the conference. “A computer probably will be able to write something like a legal opinion — maybe in 10 years. That opinion might be able to say whether state abortion bans are lawful under current doctrine, but it cannot engage in a political discourse about whether there should be a constitutionally protected right to reproductive choice.”

Livermore’s keynote will be part of a specialized legal language workshop within the 2022 Conference on Empirical Methods in Natural Language Processing, a gathering of experts in the field. If you’ve interacted with a customer service chatbot, your life has been touched by this work.

Over the past 40 years, developments in natural language processing have made the lives of lawyers everywhere much easier, as the search for relevant legal precedent has evolved from hunting through dusty books and supplements to stringing together complex Boolean connectors to simply typing out a query in plain English. (“Plain English” by lawyer standards, anyway.)

The next step might be to teach the artificial intelligence models how to navigate and apply different sources of law, from the Constitution down to local ordinances.

“It’s all about moving from a system where you need specialized expertise to navigate the legal system to one where a regular person using natural language can make inquiries and get responses back,” Livermore said.

Livermore and other computational experts are now seeking to apply their empirical analysis of legal texts to achieve other purposes, such as being able to predict the outcomes of cases or understand how language evolves in the courts over time. Livermore sees this group as part of a broader movement of legal scholars interested in empirically understanding the law. (In November, the Law School and its Center for Empirical Studies in Law hosted a conference of more than 200 empirical legal scholars from around the world.)

Livermore acknowledged that his audience in Abu Dhabi will be unlikely to include many legal practitioners. Instead, his primary message for the computational experts who work in the legal domain is to pay careful attention to the algorithms and models they are building.

Empirical models are only as good as the quantity and quality of data points they include. In a complicated and dynamic area like the law, it is difficult to find enough cases to build accurate models, particularly given the prevalence of unpublished opinions that may or may not have precedential value, Livermore said. Plus, the reasoning provided in a judge’s opinion may not accurately reflect the judge’s underlying rationale, he said.

And while it might be valuable to offer businesses an algorithm to predict the outcome of a Supreme Court case, Livermore sees all kinds of warning flags.

“Supreme Court decisions are set up to be hard to predict because the court takes up the hard cases, where the law is unclear or maybe there’s a circuit split,” Livermore said. “Even more importantly, AI models rely on large data sets to build accurate predictions, and there just are not that many Supreme Court cases.”

For young lawyers watching the industry’s evolution toward AI, Livermore offers this advice: “If you’re the kind of practitioner whose value-add is doing simple searches of legal databases, then the changes are not good for you. But if your value-add is being creative and framing arguments and understanding business contexts — higher-level tasks — then it’s great for you because you can spend your time higher up the value chain.”

“I think it just depends on what kind of lawyer you are, but I would think for UVA Law graduates, it’s generally going to be good news,” Livermore said.

Livermore is director of the Program in Law, Communities and the Environment at UVA Law, and is affiliated with several other centers, including the LawTech Center, the Center for Public Law and Political Economy, and the Center for Empirical Studies in Law. He is the host of the podcast “Free Range with Mike Livermore.”