Chatty art labels: Evaluation of a conversational museum guide

This study explores the potential of a large language model (LLM)-powered Conversational Agent (CA) as a virtual museum guide to enhance visitor experi-ences. A prototype was developed, enabling natural language interactions with an LLM-powered CA about augmented reality (AR) artwork viewed through a head-mounted device (HMD). An exploratory field study with 22 museum vis-itors found positive evaluations based on a multimedia guide assessment scale. Results showed that the most prevalent were fact-based questions, while qual-itative findings highlighted appreciation for personalized information, concerns over information gaps, challenges in formulating questions, and discomfort with public voice interactions. The study underscores the potential for LLM-powered guides for museum experiences, while identifying design challenges for future re-search and development.