Mapping Legal Propositions in the U.S. Supreme Court Party Briefs: A Human-In-The-Loop Approach

United States Supreme Court briefs influence the Court’s decisions by articulating legal propositions that provide a basis for judicial reasoning and debate. Analyzing the briefs requires substantial effort from legal professionals, and large language models (LLMs) can improve efficiency in this process. To assist legal experts in analyzing the briefs, we propose an approach that uses an open-source 8B LLM to identify legal propositions, followed by expert review and revision. We analyzed the party briefs from five First Amendment cases decided in 2025. Our results show that the language model evaluated in this study identified 72.4% of legal propositions that could be accepted directly by the legal expert, indicating that these propositions did not require formulation from scratch. Nevertheless, the model still missed 20.5% of legal propositions that the expert believed should be included. Our approach can shift expert effort from initial identification toward verification and revision. Future work may apply the legal propositions generated by this approach to subsequent stages of Supreme Court case analysis, such as the analysis of oral arguments.