Mining Coevolutionary Patterns of Opinions and Behaviors in Multiplex Social Networks
Main article
Abstract
In real-world social systems, individuals form heterogeneous opinions on the same issue and continually adjust both their attitudes and overt behaviors through repeated interaction with peers. As online social platforms become deeply interwoven with offline life, the interplay between opinion diffusion and behavioral choice has grown more pronounced, and a single-layer perspective is no longer sufficient to capture the true trajectory of collective evolution. Motivated by this gap, this article studies the coevolution of opinions and behaviors over a multiplex (two-layer) social network and develops a unified data-mining framework for discovering coevolutionary patterns and stable regimes. In the baseline specification, the opinion layer follows a DeGroot-style weighted update rule, while the behavior layer integrates three influence channels operating in parallel: neighbor imitation, payoff-driven adaptation, and cognition-behavior consistency. Around this model, we first investigate the synchronous-update regime and characterize the joint evolution of the average opinion and the cooperator share, then study the asynchronous-update counterpart and compare the two regimes in terms of convergence speed, transient pathways, and steady-state outcomes. Three canonical topologies are examined: random graphs, small-world graphs, and scale-free graphs. Finally, the model is calibrated against a longitudinal household survey on pro-environmental attitudes and recycling behaviors. Results show that opinions and behaviors coevolve along strongly coupled trajectories; that the opinion-dependency parameter, the imitation weight, and the payoff sensitivity exert decisive influence on consensus formation and on the diffusion of cooperative action; that in most regimes the cooperator share stabilises before the average opinion does, indicating a behavior-leads-cognition-lags pattern; and that scale-free and random networks tend to converge faster and to higher steady cooperator levels than small-world graphs. Synchronous and asynchronous mechanisms agree on long-run outcomes but differ markedly along the transient path. The simulated trends match the survey data within 4 percentage points, confirming that the proposed framework offers a practical and explanatory tool for mining coevolutionary patterns in real multiplex social systems.
