The Meta-Body of Artworks: Sequence-Based AI Storytelling for Culturally Situated Heritage Narratives

Abstract

This paper develops the concept of the meta-body of artworks as an evolving relational field through which artworks acquire meaning across exhibitions, narrative sequences, curatorial framings and interpretive encounters. Rather than treating artworks as isolated digital objects or AI as an autonomous narrative author, it proposes a model of sequence-based AI storytelling in which cultural meaning is organised through governed relational structures. The paper makes four connected contributions. Conceptually, it reconceives digitised artworks as components of relational semantic infrastructure. Empirically, it presents exploratory findings from a European University Institute pilot. Architecturally, it proposes a blueprint based on relational graphs, multimodal embeddings and constraint-based traversal. From a governance perspective, it embeds community authority, metadata and refusal mechanisms within the structure of the system rather than treating them as external policy layers. The proposed meta-body architecture treats provenance, curatorial framing, sequence, motif relations and accumulated interpretation as parts of a persistent semantic infrastructure. Within this framework, AI assists traversal through culturally bounded possibility spaces while interpretive authority remains with users, curators and communities. The empirical component draws on a five-session RIZOM pilot with 43 participants at the European University Institute. The study examined how artwork presence, abstraction and contextual anchoring affected reflective and symbolic traversal. Results were interpreted descriptively and directionally rather than causally. The findings suggest that symbolic practice can stabilise through repeated use, that excessive conceptual and visual abstraction can reduce participation and interpretive trust, and that representational anchors can support recovery following symbolic overload. The study provides directional evidence for the conditional carrying capacity of symbolic fields while leaving implementation of the full semantic infrastructure for future work. The paper further develops design principles for sequence-first narrative systems, metadata as contextual boundary, AI as facilitator, calibrated visual anchoring, governed co-construction and refusal. It positions narrative AI as cultural infrastructure supporting community expression and intercultural encounter while preserving contextual and relational integrity. This paper was accepted and presented at AI4CHIEF'2026, the Symposium on Artificial Intelligence for Cultural Heritage and Indigenous Futures, CNAM, Paris, April 2026.