摘要：AbstractMathematical modeling plays a fundamental role in understanding how social influence shapes individuals’ opinions. Although most opinion dynamics models assume that individuals update their opinions by averaging others’ opinions, we point out that the weighted-averaging mechanism features a non-negligible unrealistic implication. We propose a new micro-foundation of opinion dynamics, i.e., the weighted-median mechanism, in the framework of cognitive dissonance theory and resolves the shortcomings of weighted averaging. Validation via empirical data indicates that the weighted-median mechanism significantly outperforms the weighted-averaging mechanism in predicting individual opinion shifts. Compared with the averaging-based opinion dynamics, the weighted-median model, despite its simplicity in form, replicates more realistic features of opinion dynamics, and exhibits richer phase-transition behavior depending on more delicate and robust network structures. The novel weighted-median model opens up a new line of research and renovates our understanding of the opinion formation process.
关键词：KeywordsOpinion dynamicsSocial networksMulti-agent systems
