Understanding audience engagement in AI-generated biophilic architecture: a multi-criteria analysis using the ELECTRE model inspired by traditional Omani design

Abstract

Building on Kellert’s framework of biophilic architecture, AI-generated scenes inspired by traditional Omani architecture and associated student engagement were examined. A visual content analysis of 122 photographs from the Al Hafa District identified recurrent biophilic attributes. These features informed the generation of 66 Sora text-to-video scenes; representative frames, produced using fixed-structure prompts to constrain variation, were posted to Instagram. Within a cohort of 200 Omani students, per-image “likes” were recorded as an engagement indicator and used to shortlist images for expert review. In parallel, semi-structured interviews with students and domain experts were analyzed via reflexive thematic analysis. The ten most engaged images were rated by five experts on eight biophilic criteria. An ELECTRE procedure synthesized the criteria to derive dominance relationships and a final ranking. Three images achieved the highest dominance and exhibited biophilic richness (e.g., vegetation, biomorphic patterning, soft daylight, and prospect–refuge). Engagement partially aligned with expert assessment: several highly liked images ranked lower in richness, indicating that “likes” may reflect visibility-weighted, low-friction behavior rather than deeper affective or cognitive responses. This study triangulates AI imagery, platform analytics, expert multi-criteria evaluation, and qualitative interpretation, offering a culturally specific approach that links engagement with biophilic attributes.