Relational Narrative Historiography: A Higher-Order Framework for Representing Competing Historical Narratives

This paper presents Relational Narrative Historiography (RNH) as a framework paper and testable research hypothesis, not as an empirical study. Historical accounts of the same event frequently diverge across archives, languages, and interpretive communities—a challenge that extends to any domain where reports conflict or facts remain disputed. Conventional knowledge graphs and Large Language Models may favour frequently represented or dominant accounts, potentially reducing the visibility of minority perspectives and of the uncertainty inherent in contested information. This paper proposes that narratives—or more generally, contested information bundles—can be represented as higher-order objects that bind events, claims, actors, sources, locations, temporal anchors, interpretive frames, and internal relations. We introduce Relational Narrative Historiography (RNH), a hypergraph-based framework in which narratives are first-class objects. Unlike pairwise graph models, which represent a narrative through separate narrative nodes and edges, hypergraphs can keep the membership of a narrative explicit and represent higher-order relationships as a single structure. The framework uses three epistemic constructs: Epistemic Weight (the strength of evidence supporting a narrative), Divergence (the quantitative distance between narratives), and Uncertainty (what remains unresolved). We describe the conceptual architecture and mathematical formulation, illustrate the calculations with a numerical example based on semantic similarity, and outline an evaluation protocol. All numerical values presented in the worked example are illustrative and are not derived from any expert-annotated real-world dataset. RNH is proposed as a framework and testable research hypothesis, but its validity still needs to be examined empirically in digital humanities, AI, and knowledge representation.