CRISIS COMMUNICATION ON SOCIAL MEDIA: POSSIBILISTIC REASONING OVER INCOMPLETE EVIDENCE

This study presents a mathematically grounded framework for decision-making when digital reports are fragmentary, delayed, or conflicting. We formulate an ordinal approach in which social signals are mapped to fuzzy predicates and encoded as constraints that induce least-committal distributions. Heterogeneous cues are combined with t-norm/t-conorm operators (default max–min), and time evolution is handled by a max–min convolutional filter with bounded drift. Information spread and credibility are modeled on interaction graphs via a trust-weighted diffusion operator that is monotone and admits a least fixed point. Operational triggers are cast as threshold rules on dual statistics necessity for conservative guarantees and possibility for coverage supported by complexity-aware algorithms and theoretical properties (monotonicity, non-expansiveness, and fixed-point existence). Experimental design uses event-wise splits, ordinal metrics (guaranteed precision vs. alert rate, coverage width), and ablations over fusion families, kernel bandwidth, and trust decomposition. A flood-onset case study demonstrates timely alerts under missing cues, while a public-health rumor scenario shows transparent containment dynamics and rapid convergence. Across scenarios, necessity-based alerts achieve high guaranteed precision at controllable alert rates, with tunable lead time and auditable contributions from cues, rules, and trust links. The framework provides a practical substrate for ethically governed operations, and a foundation for future hybrid probabilistic-ordinal models and learned encoders.