Profiling phubbing practitioners using machine learning

Phubbing is a heterogeneous digital behavior that may occur within different configurations of personality and problematic social media use. This study identified multivariate profiles of Brazilian adults on the basis of Big Five personality traits, problematic Instagram use, and the communication disturbance and phone obsession dimensions of phubbing. The initial sample included 1,551 adults; the primary analysis included 1,470 current Instagram users. Eight standardized indicators were analyzed using Gaussian Mixture Models with one to six components and alternative covariance structures. Model selection prioritized the Bayesian information criterion (BIC), with the Akaike information criterion (AIC) used as complementary evidence. The five-component diagonal model had the lowest BIC, whereas the AIC favored a six-component diagonal solution. The BIC-preferred model identified five partially overlapping profiles: Regulated Low-Digital (16.9%), Introverted Low-Digital (15.6%), Moderate Digital (21.1%), Socially Engaged Digital (20.8%), and Digitally Vulnerable (25.6%). The average maximum posterior membership probability was 0.803, indicating meaningful classification uncertainty. The profiles differed most strongly in problematic Instagram use, Phone Obsession, Communication Disturbance, Conscientiousness, and Extraversion. The findings indicate that phubbing-related behaviors occur within heterogeneous personality-digital configurations rather than a single psychological pattern. The profiles should be interpreted as probabilistic and descriptive, not as fixed psychological types or diagnostic categories. Longitudinal and independent replication studies are needed to examine their stability and practical relevance.