Speakers
Description
Definitions play a crucial role in the acquisition, transfer, or translation of specialized knowledge. In the legal profession, precise definitions of complex concepts facilitate seamless knowledge management, application, and interpretation. However, despite growing reliance on artificial intelligence (AI) for this task, the real-world feasibility and user acceptability of AI-generated legal drafting remain largely unexamined. When addressing this gap, it is critical to consider that the use of authoritative legal terminology is central to the validity of any legal definition. Indeed, precise terms facilitate the application and interpretation of core concepts. However, new legal domains lack stable legal terms and conceptually demarcated concepts. Resolving the inherent tension between linguistic evolution and the need for legal certainty make the legal domain a unique use case that necessitates domain-specific task adaptation for general-purpose large language models (LLMs). With that in mind, this paper explores the potential of LLMs to generate legal definitions perceived as clear, precise, and usable by domain users through an exploratory study of a generated dataset of key influencer-related concepts. The results underscore the critical importance of data quality and precise prompt engineering when leveraging LLMs for complex, domain-specific tasks.