Nonlinear polarization modeling via dual-threshold diffusion–evolution dynamics

For decades, researchers have sought to understand how human opinions evolve. In today’s fragmented social-media landscape, the creation and diffusion of public opinion shift instantaneously, increasing the risk of global polarization. Existing models usually focus on a single psychological mechanism and therefore struggle to link individual cognition with macro-level evolution. We propose an opinion-evolution dynamics framework that couples diffusion models with evolutionary algorithms and is driven by a dual-threshold social-cognition mechanism. In the diffusion stage the framework continuously generates diverse opinions and, guided by a consensus-fitness function, highlights mainstream high-value views; in the denoising stage evolutionary selection preserves the fittest opinions and injects controllable perturbations to break information bubbles. Experiments on four classical network topologies reveal nonlinear polarization patterns. This paper uncovers the complex dynamics of opinion evolution, deepens understanding of how social cognition and network structure shape opinion dynamics, and provides a broader perspective for managing and guiding the evolution of public sentiment.