Everyday Instagram behavior maps distinct facets of loneliness in young adults

Abstract Loneliness is a major health risk and is often hard to detect. It is multidimensional and can involve social, emotional, and structural facets with different causes and behavioral consequences. An open question is whether these facets can be differentiated in everyday Instagram behavior. We analyzed consented Instagram data exports from 415 US college students and modeled five validated measures related to loneliness using activity patterns, language, and images. Multimodal traces distinguished subjective loneliness but were less informative for structural isolation. Model explanation analyses revealed dissociable signatures: social loneliness was linked to low interpersonal engagement (fewer likes and comments), whereas emotional loneliness was linked to negative affect (negative language and sad/angry facial expressions) and late-night activity. Together, these results provide a cross-measure behavioral “construct map” of loneliness, suggesting that multimodal digital traces can separate markers of dissatisfaction with social connections from markers of emotional isolation, moving beyond generic prediction toward construct-sensitive characterization. Because the sample was of college age, predominantly female, and largely Asian or White, an important next step is to test whether these signatures generalize to more diverse populations.