Mind the model: Understanding the role of peer influence in population responses to epidemics

Abstract Mathematical modelling and computer simulation have become essential tools for understanding infectious disease dynamics. The COVID-19 pandemic underscored the strong interconnection between human behaviour and disease progression. For instance, the adoption of protective behaviours-such as mask-wearing and social distancing-curbs transmission, while a worsening epidemic can drive greater adoption of such behaviours. This mutual influence calls for models that incorporate human behaviour endogenously. Importantly, these models must be grounded in real-world data, both epidemiological surveillance data and data on population behaviour, which are increasingly available. In this study, we develop an age-stratified SIR (Susceptible-Infected-Recovered) compartmental model, further stratifying individuals by their engagement in protective behaviours (e.g., mask usage). Behavioural change is prompted not only by epidemic trends but also by peer influence. To capture this dynamic, we integrate data from the IMMUNE survey (Modelling the IMpact of huMan behavioUrs on iNfections sprEad, ERC grant), which investigates patterns of interpersonal trust and communication across social groups. Based on these data, we construct discussion matrices-analogous to contact matrices in traditional models-capturing the frequency and relevance of discussions (in person, online, or by phone) across age groups and social spheres (cohabitants, family, friends, schoolmates, coworkers). We then simulate scenarios where behavioural contagion spreads either uniformly across the population or structured by these matrices. Our results show that peer influence significantly shapes behavioural responses, which in turn affect epidemic trajectories. This study offers an innovative, data-informed modelling framework for integrating social communication dynamics into infectious disease models.