Abstract
Loneliness, a pervasive mental health concern, is often misconstrued as an individual pathology, limiting our understanding of social effects via peer-to-peer interactions. This study investigates how homophily (similarity-based connectivity) and induction (interacting mental states) contribute to loneliness clustering. Using a computational model, we simulate social network interactions via established induction frameworks: emotional, behavioral, and cognitive contagion. We map these pathways to fundamental processes of simple contagion, complex contagion, and self-activation, explaining how ideas and behaviors spread. Results show that high homophily is necessary for loneliness clustering, and the model recovers empirical findings of network clustering (a positive correlation of individuals' mental states beyond direct neighbors) with extended "degrees of influence" across networks and setups. This universality of clustering across pathways renders the metric uninformative in screening causal mechanisms behind loneliness clustering. Fortunately, each inductive pathway displays distinct out-of-equilibrium dynamics, aiding in identifying real-world mechanisms. The study emphasizes the significant role of individuals' social contexts in loneliness and calls for a shift from static to dynamic measurements in loneliness research. This shift will enhance the relevance of future research on evolutionary patterns in real-world social network data, leading to a more robust understanding of the mechanisms of loneliness.