Exploring Shared and Unique Predictors of Positive and Negative Risk-Taking Behaviors Among Chinese Adolescents Through Machine-Learning Approaches: Discovering Gender and Age Variations.

Despite extensive research on the impact of individual and environmental factors on negative risk-taking behaviors, the understanding of these factors' influence on positive risk-taking, and how it compares to negative risk taking, remains limited. This research employed machine-learning techniques to identify shared and unique predictors across individual, family, and peer domains. Participants (N = 1012; 44% girls; Mage = 14.60 years, SD = 1.16 years) were drawn from three public middle schools in a large city in southern China (with 49.2% in grade 7 and 50.8% in grade 11). The findings indicate that positive risk-taking is significantly associated with general risk propensity, self-control, and negative parenting by father, while negative risk-taking is correlated with self-control, deviant peer affiliations, and peer victimization. Paternal negative parenting triggered positive risk-taking in boys, whereas self-control had a greater impact on girls. For negative risk-taking, boys were more affected by peer victimization, while girls were more influenced by deviant peer affiliations. This study further demonstrates that as progress from junior to senior high school, peer influence grows more significant in predicting positive risk taking; deviant peer affiliations exert a persistent pivotal influence, future positive time perspective replaces life satisfaction, and paternal negative parenting becomes increasingly impactful in predicting negative risk taking.