Aniket Kamat Rose Hills
Deep Learning for Causal Inference in Legal Settings
Two of the biggest challenges in legal research are conforming to privacy constraints in real-world experiments and evaluating causal effects in high-dimensional settings. When multiple people are working on a single project, the former often requires physically mailing any personally identifiable information (PII), slowing down collaboration and risking data falling into the wrong hands. High-dimensional datasets also make traditional causal inference more difficult due to many factors including the increased potential for multicollinearity and the increased sample size required to estimate the causal effect.
Federated and deep learning methods can be used to address both of these problems. Using federated learning can greatly reduce the time it takes to collaborate on models that require PII data while deep learning methods can help preserve causal structures in complex datasets and minimize data exposure. My project aims to develop software using these tools to improve causal inference under legal data constraints, with the goals of contributing to the growing field of causal deep learning and promoting the practical application of data-driven decision-making to legal policy.
Message To Sponsor
Thank you for your generous donation! I'm really grateful for the opportunity to work on this project this summer. I've been interested in using statistics for legal research for most of college, so I sincerely appreciate the opportunity to fully explore that interest.