Designing an AI Agent to Facilitate Reflective Art Appreciation Through Feldman’s Critique Framework

다국어 초록 (Multilingual Abstract)

Interest in art appreciation has grown in recent years, yet many individuals—especially those without prior knowledge of art—still find it difficult to interpret and engage with artworks. Feldman’s art critique method, one of the most representative art critique frameworks, provides a structured and systematic art appreciation, but it still remains challenging for non-experts to apply in real-world appreciation contexts. While interactive approaches have been introduced to support art appreciation, most of them still focus on one-directional information delivery, and research on technologies that explore viewers' deeper engagement with art appreciation remains limited. Thus, we propose an AI-based agent that integrates Feldman’s art critique stages into an interactive web-based system for guided art appreciation. To achieve this, a large language model (LLM) integrated with a lightweight RAG pipeline was applied, and custom datasets were created for the RAG. A preliminary study with 10 individuals was conducted to explore user expectations for AI-supported art appreciation. Most participants expressed a preference for interactive systems that provide clear and informative guidance. Based on these insights, the proposed system was designed with the goal of supporting step-by-step interpretation, and contextual understanding of artworks. An exploratory user study with two participants was conducted, revealing positive responses toward the AI-assisted Feldman art critique system.