Improving Advocacy Resource Allocation through Predictive Legislation Tracking

Partners(s): American Civil Liberties Union (ACLU)
Status: Paused
Github Repo: https://github.com/dssg/state-leg-tracker
Team: Kasun Amarasinghe, Liliana Millan Nunez, Kit T. Rodolfa, Rayid Ghani

We developed a machine learning system for tracking state-level legislation. Given a legislative bill the ML system, (1) predicts the likelihood of that bill being passed into law, and (2) classifies the bill into one or more of ACLU’s topic areas. We built these models based on publicly available data (in partnership with LegiScan) and have publicized the code repository on GitHub. For advocacy organizations like the ACLU, tracking state-level legislation is a necessary step to identify where to focus and prioritize their advocacy resources. Predicting passage of a legislative bill can be difficult for humans, even for experts at advocacy organizations, as success of a bill depends on a range of different factors. Assigning legislative bills to their respective topic areas, while easier than predicting passage for a domain expert, often requires significant manual efforts to keep up with legislation. These manual effort demands often result in a trade-off between recall (out of all important legislation, percentage of identified important legislation) and efficiency (time taken to identify important legislation) of the process. A data-driven legislation tracking system that predicts passage likelihood, and classifies bills into their topic areas can help advocates identify important legislation improving both recall and efficiency of the process.