Abstract
In an era where digital communication accelerates the global spread of false narratives, understanding how misinformation and disinformation propagate, especially during crises such as the COVID-19 pandemic, is vital to public health and policy. To delve into the diffusion mechanisms of misinformation (unintentionally false information) and disinformation (intentionally false information), we introduce a novel enhanced agent-based model (ABM) that integrates psycho-social factors and communication networks, which are elements often overlooked in traditional equation-based models (EBMs). We assess the two distinct techniques (ABMs and EBMs) through the lens of six classical SEIRS-class models (S susceptible, E exposed, I infected, R recovered). Beside the enhanced ABM, we also develop a simple ABM to emulate the EBM structure. We compare the ABMs with the EBMs over their entire parameter ranges in a total of 11110 experiments. Results show an overall weak equivalence between the two types of models, even if, under certain conditions, the outcomes of the EBMs and ABMs are similar. Furthermore, we evaluate the two model types by fitting them to real-world infodemic data on vaccine acceptance over 36 weeks using a multi-objective optimization procedure. The enhanced ABM shows an exceptionally better fit to real-world data (Pearson's correlation coefficient [Formula: see text] and normalized root mean of square error NRMSE = 0.055) than the EBM ([Formula: see text], NRMSE = 0.418) and the simple ABM ([Formula: see text], NRMSE = 0.103). These findings underscore the critical role of model structure in capturing infodemic dynamics, and advocate for the use of ABMs when psycho-social influences and network interactions are central to the phenomenon.