The Risk of Bullying and Probability of Help-Seeking Behaviors in School Children: A Bayesian Network Analysis

An increase in aggressive behaviors in adolescents has been observed for a few years. The participation in bullying is associated with many psychosocial difficulties in adolescent development. On the other hand, the help-seeking behavior can be one of the most important protective factors that reduce the risk for this type of violence. The study was aimed at estimating the risk factors, as well as the protective factors of school bullying, by using the Bayesian networks to build a model allowing to estimate the probability of occurrence of the aggressive and help-seeking behaviors among school children. The focus was on individual risk/protective factors related to EAS temperament (emotionality, activity, and sociability) and variables related to the family context (level of cohesion, flexibility, family communication, and family life satisfaction). Bayesian methods have not been particularly mainstream in the social and medical sciences. The sample comprised 75 students (32 boys and 43 girls), aged 13–15 (M = 13.82; SD = 0.47). Assessment comprised The EAS Temperament Questionnaire, Family Adaptability & Cohesion Evaluation Scales FACES IV-SOR (Family Rating Scale), and Survey questionnaire. The Bayesian networks were applied. Depending on the values of the identified variables, very high a posteriori probability of bullying and help-seeking behaviors can be predicted. Four EAS subscales (Distress, Fear, Activity, Sociability) and two SOR subscales (Balanced Flexibility and Balanced Cohesion) were identified as predictors of bullying. Moreover, two SOR subscales (Family Communication and Life Family Satisfaction) and one EAS subscale (Sociability) were identified as predictors of help-seeking behaviors. The constructed network made it possible to show the influence of variables related to temperament and variables related to the family environment on the probability of bullying or the probability of seeking help and support. The Bayesian network model used in this study may be used in clinical practice.