UC Berkeley

The present work, titled “Essays on Law and Natural Language Processing”, explores the uses of Natural Language Processing (NLP) in law. Structurally, the first two essays explore the applications of NLP methods in the legal domain. The third essay explores the epistemological issues concerning NLP applications in the legal domain, particularly in the context of the possibilities of using large language model (LLM)-based artificial intelligence (AI) for legal decision-making. Accordingly, the first essay proposes the application of text classification methodology to the question of US Supreme Court’s certiorari petition outcome prediction. The second essay proposes the application of word embedding-based synchronic lexical semantic change detection methodology to investigate a question in corporate governance regarding corporate honesty on Environmental, Social and Governance (ESG) issues. The third essay argues that while large language models are not explicitly designed to capture reasoning, they will likely exhibit latent legal epistemological biases rooted in Jean-Michel Berthelot’s schemas of intelligibility-based patterns of legal reasoning detected in the text. Overall, this dissertation argues that the complex nature of legal reasoning and the resulting textual data generation necessitate not only a cautious approach to the application of NLP methods in law but also a re-examination of the epistemological foundations underpinning legal reasoning.