Adapting RAG to Specialised Communities: A UX-Driven Design Study
(2026) Human-AI Interaction and Experience Design (HAXD 2026) — Location: Valencia (9.June.2026)
Files
2026173866.pdf
Open Access - Adobe PDF
- 184.69 KB
Details
- Authors
- Abstract
- Large language models (LLMs) substantially transform information retrieval, with retrieval-augmented generation (RAG) emerging as a predominant paradigm. Although RAG demonstrates strong performance for generic information needs, its adaptation to specialized user communities remains insufficiently understood. In this paper, we address this issue through a collaboration with the Académie royale de Belgique, focusing on the design of a RAG-based search engine for the Biographie nationale de Belgique, a scholarly biographical corpus that serves as the system's exclusive knowledge base. We conduct contextual inquiries with members of the academia and develop user experience (UX) artifacts, including personas and a consolidated sequence, to elicit information needs, identify breakdowns in the tasks, and derive system requirements. Our results indicate that UX research methodologies offer a robust and systematic foundation for characterising the design constraints required to tailor a generic RAG system to the needs of a specialised user community.
- Affiliations
Citations
Rolin, E., Watrin, P., & Kieffer, S. (2026, June 5). Adapting RAG to Specialised Communities: A UX-Driven Design Study. Human-AI Interaction and Experience Design (HAXD 2026), Valencia. https://hdl.handle.net/2078.5/278599