As I and many others have covered, the rapid infusion of new technologies into law—what some refer to as “law+tech”—is one of the major transformational trends leading to the post-normal era in which lawyers find themselves. But there is a very broad spectrum of law+tech initiatives coming into play, from those automating quite mundane routinized processes to those in pursuit of what I would call the Holy Grail of law+tech—predictive analytics.
Those who follow the Computational Legal Studies blog are familiar with the powerful predictive analytics tools Dan Katz and Mike Bommarito are developing for law, most notably their work on Supreme Court decisions. Over here at the Vanderbilt Program on Law and Innovation, John Nay, a Vandy Engineering Ph.D. Candidate and PoLI Research Fellow, is also developing tools for predictive legal analytics, in his case on federal legislation. I’ll let John’s words explain what’s behind the project:
While working at a policy strategy firm in D.C. and while interning for the Majority Leader of the U.S. House, I was overwhelmed with the number of bills to track. After leaving D.C. and no longer reading Politico every morning, trying to keep up-to-date was hopeless. There are often more than 8,000 bills under consideration in Congress but less than 4% are likely to become law. Based on my research on predicting and understanding legislation with natural language processing, I created a machine learning system to predict bill enactment. Starting with the 107th Congress, models were trained on data from previous Congresses, and all bills in the current Congress were predicted until the 113th Congress served as the test. The median of the model’s predicted probabilities for enacted bills was 0.71, and the median of the predicted probabilities for failed bills was 0.01. To bring this predictive power to the public, I built a web interface, PredictGov, where all bills currently under consideration and their predictions (updated daily) can be interactively explored. Users can sort and filter the bills and download the results. I also provide an application for searching networks of similar bills based on their texts on the website and updates on key bills on Twitter @PredictGov.
I’m delighted to be working with John to help inform his project and other initiatives, even though I understand only half of what he’s talking about! Look for more to follow on John’s PredictGov website.