In New Book, Michael Livermore Delves Into Data’s Possibilities

Professor Michael Livermore of the University of Virginia School of Law has collaborated on a new book that examines how computer-based tools are being used to analyze dense legal information — everything from proposed statehouse bills, to parole hearings, to the collective writings of the U.S. Supreme Court.

Livermore co-edited “Law as Data: Computation, Text, and the Future of Legal Analysis” with Daniel Rockmore, a professor of computer science and mathematics at Dartmouth College. Their book was released by Santa Fe Institute Press.

The book is a product of a relatively new research movement, and Livermore has been among the researchers on the forefront worldwide.

“The book developed from a workshop held at the Sante Fe Institute in 2017, where leading scholars in law, social sciences, machine learning and mathematics met to discuss cutting edge research using legal texts as data,” Livermore said. “It was a wonderful interdisciplinary group working on a wide range of questions for many different perspectives — and it made for an incredibly interesting two days of conversation. Dan [Rockmore] and I decided to put together this book as a way to keep the conversation going and introduce a broader audience to these ideas.”

The work collected in “Text as Data” shares a common approach: translating the texts in legal documents into data and then using mathematical tools to make sense of the resulting information.

In one example from the book, Livermore and his research team used “sentiment analysis” to see how the collective writing style of the U.S. Supreme Court has changed over time.

“Sentiment analysis is a tool that basically looks at a written text, extracts out negative versus positive words — so positive words are like ‘happy’ and ‘good’ and ‘wonderful,’ and negative words are like ‘bad’ and ‘terrible’ — to can get a sense of the general attitude this text offers the world,” Livermore said.

Using the tool, he found that the Supreme Court’s language has grown distinctly more negative over time.

“So we say the Supreme Court has gotten grumpier over its 230-year history,” he said.

Another example of how the technology is being used is to scan reams of information generated by the policymaking process to predict outcomes, such as whether a bill will become a law. Researchers with the company FiscalNote, which focuses on government data analytics, report in the book results from their state-level research.

“The FiscalNote team has built a strong predictive model that does well across all 50 states,” Livermore said. “It’s an approach you could never imagine doing without this kind of data, without these kinds of new analytic tools that exist these days.”

Livermore has conducted separate research with Fiscal Note about the public comments regulatory agencies receive. Agencies have been required since the 1940s to collect public input on certain decisions, and Livermore said text analysis tools can help provide context amid the volume. He said the new technology may also be able to root out “comment bots,” which submit fake comments under real people’s names.

Additional chapters in the book cover the types of language associated with successful parole hearings in California; the influence of French-language jurisdiction on the European Court of Justice; the use of information on cross-references in statutes to detect legal influence; and a predictive model of employment law cases in U.S. district courts.

Livermore joined the faculty in 2013. He teaches a class called LawTech that covers the importance of technology for the legal profession. His other classes include environmental law, administrative law and regulatory law and policy.