Javascript must be enabled to continue!
Supporting UI designer's decisions with AI RAG assistance

Decisions in interface design are usually based on standards, style guides and books that designers consult frequently. Conversational Large Language Model (LLM) assistants could speed up this process, but these documents often cannot be attached to them, because licensing terms and company policies forbid their use in third-party services. This work studies the feasibility of adding a designer's reference library to a local LLM through Retrieval-Augmented Generation (RAG), so that each answer is grounded in personalized content and cites its source. Four conditions were compared: the local model alone, the same model with text retrieval, the same model with visual retrieval, and a commercial assistant as reference.

In order to evaluate these conditions, 48 questions based on a designer’s professional practice were formulated and applied to the four conditions. An expert designer then conducted a blind evaluation of the responses. The results of the experiment show that the use of retrieval did not increase the number of resolved questions, since text retrieval and the commercial assistant solved practically the same number of questions as the local model alone. However, it did make the answers verifiable, as 42 of the 48 answers were confirmed against the cited page, with a citation precision of 0.94. Given that the corpus is stored in the designer's own machine, each citation links to a page that can be opened during the search itself, and the material is not shared with outside entities.

These preliminary results suggest that local assistants based on sources can be a responsible way to introduce AI in design practice, respecting licensing restrictions and allowing designers to keep control over their references.

Attachment language: English File type: PDF document Presentation Poster
Updated: 02-11-2026 16:39 - Size: 507.93 KB
Attachment language: English File type: PDF document Presentation File
Updated: 02-11-2026 16:40 - Size: 6.1 MB
Speakers
Marina MaciáDavid Rizo Valero

Back

ACM GreekCHI-CYCHI Symposium 2026


We are excited to invite you to the ACM GreekCHI-CYCHI Symposium 2026, held on November 6-7, 2026 jointly in Athens, Thessaloniki, Nicosia and Online.