Analysis of Color Expression and Emotional Keyword Perception Characteristics in Generative AI Spatial Images Based on Wood Learning

This study investigates how undergraduate students’ material perception of wood is structured through color expression and emotional language in AI-mediated spatial visualization. Thirty-two spatial images were generated by second-year interior architecture students using AI following a lecture on wood materiality. Participants selected target wood species and emotional keywords, then submitted descriptive statements documenting their material selection intent. RGB values extracted from wood-dominant regions were converted to CIELAB coordinates (L*, a*, b*) for quantitative analysis, and integrated with emotional vocabulary data through a mixed-method approach. Results indicate that wood color representations were concentrated in the mid-lightness range (L* 30–50, 65.6%), with all cases distributed in the positive chromatic quadrant, confirming consistent perception of wood as a warm-toned, yellow-dominant material. K-means cluster analysis identified three typological groups — high-chroma mid-lightness, mid-chroma mid-lightness, and low-chroma low-lightness — each systematically corresponding to distinct emotional vocabulary patterns. Cross-analysis confirmed that material perception formed through theoretical learning is consistently externalized via prompt verbalization and structured into identifiable color typologies through AI-mediated visualization. These findings suggest that generative AI functions as a mediating system that translates learners’ internalized material perception into visual color expression, providing empirical foundations for emotion-centered design education frameworks.