Theory-driven LLM-assisted analysis in crisis communication: A case study of the self-oriented model of digital publics

초록

This study introduces a theory-driven, human-centered LLM workflow for scaling interpretive constructs in public relations research, using the Self-Oriented Model of Digital Publics (SOMP; Lee, 2023) as an illustrative case study. Applying this workflow to 3072 crisis-related tweets from the CrisisLexT26 dataset, we demonstrate how survey-based perceptual grounding, LLM-based classification, and behavioral validation can be systematically integrated to validate and extend emerging public relations theories at scale. Results show strong convergence between human raters and LLM classifications and reveal that SOMP identity orientations (I-, You-, We-) systematically predict patterns of emotional expression and communicative intent across crisis origin and time. The proposed three-phase workflow advances public relations scholarship by establishing a replicable methodological pathway for applying AI as a scaling instrument for theoretical constructs in digital listening, public segmentation, and crisis communication research. © 2026 Elsevier Inc.

키워드