On Modeling Agent Behavior Change Through Multi-Typed Information Diffusion in Online Social Networks

The rapid spread of information on online social networks (OSNs) has had both beneficial and detrimental societal impacts. Notably, the propagation of misinformation during events such as the COVID-19 pandemic has driven collective behavioral responses, sometimes exacerbating crises such as panic buying and bank runs. The toilet paper shortage in Japan in 2020 caused by people’s selfish behavior owing to the diffusion of information to correct misinformation, which is considered a general phenomenon based on three sociological characteristics: a collective action problem, a self-fulfilling prophecy, and pluralistic ignorance. Understanding what kind and how information can inhibit selfish behavior to mitigate the problem under the phenomenon is critical for developing strategies to mitigate it. However, existing models addressing misinformation diffusion often lack the capacity to represent the complex behavioral phenomena driven by multiple information types and do not fully capture the subjective decision-making processes of individuals influenced by information. Therefore, this study introduces a computational model that combines subjective logic to represent opinion changes of individuals using multi-typed information and cumulative prospect theory to simulate individuals’ decision-making. This model aims to depict the effectiveness of information in inhibiting selfish behavior (behavior-guiding information) in the phenomenon. The simulation results demonstrate that behavior-guiding information can either inhibit or unintentionally promote selfish behavior, depending on network parameters and information characteristics. These findings suggest that the strategies of behavior-guiding information for this phenomenon should be carefully designed, and that our model could help in these design.