Laying Down the Law with Class Action Lawsuits
How do judges choose a fair amount for attorney’s compensation in class action lawsuits? Is the decision consistent across similar court cases? Are there quantifiable defining characteristics that form the basis for differentiation in attorneys’ fees awards? Attorney and Milton R. Underwood Chair in Free Enterprise Law Professor Brian Fitzpatrick seeks to answer these questions.
This collaboration between the Data Science Institute (DSI) and Professor Fitzpatrick began to accelerate the characterization and extraction of information from several documents in class action lawsuits. Using all of the class action lawsuits that were settled in federal court from 2008 to 2015 – more than 3,000 – and the help of many law students, Professor Fitzpatrick gathered over 18,000 documents from the dockets of these cases and characterized these lawsuits based on 55 expert-identified features.
With automation in mind, an integrated team of law students, undergraduate students from diverse educational backgrounds, and DSI data scientists examined a variety of natural language processing (NLP) techniques alongside Professor Fitzpatrick to extract the desired information from these documents. This project uses traditional NLP, machine learning, and leverages state-of-the-art deep learning techniques to discover, extract, and identify the characteristics of these class action lawsuits. Students from the DSI’s Summer Research Program have contributed a number of machine learning models to the effort and a framework for rapidly training and evaluating these models in a reproducible manner, and the data science skills gained from this project have enabled participants to enter data science-focused doctoral and graduate programs and leading data science companies.