The Trap of Incomplete Information Games in the Attention Economy: A Simulation Study of Network Collapse and Restoration Strategies for Informational Hygiene

In contemporary online environments, the dynamics of the attention economy drive payoff-seeking, irresponsible information diffusion, resulting in severe discourse polarization. This study defines the primary objective function for society as the maintenance of “Informational Hygiene.” To capture these dynamics, we propose an Adaptive Co-evolution model that integrates incomplete information games with bounded rationality and evolutionary opinion dynamics. We incorporate a subjective expected payoff function, based on information asymmetry, into a scale-free network model. The interaction between agents’ internal states (infection degree, psychological damage, resilience) and network topology changes (edge deletion and rewiring) is formulated as a hybrid dynamical system of coupled ordinary differential equations and discrete topological updates. Extended computational experiments evaluate the coupled dynamics of autonomous edge deletion (structural bypass) and global propagation suppression (circuit breakers). The simulation results demonstrate that “subtractive interventions”—creating discriminatory boundaries by completely severing ties with contamination sources—can halt information spread but tend to fragment the social network irreversibly. Depending on inherent resilience parameters, this structural collapse conditionally isolates agents from social buffering mechanisms, causing a “Paradox of Isolation” characterized by strong hysteresis, where internal damage diverges. To ensure informational hygiene and prevent irreversible polarization, we suggest that platforms must implement “reasonable accommodation.” Specifically, “additive interventions”—comprising autonomous rewiring to healthy nodes to maintain inclusive boundaries (weak ties) and foster the complex contagion of resilience—are mathematically optimal prerequisites for next-generation intervention platforms.