Bayesian Network modeling for Dark Triad, aggression, and empathy

This study aimed to construct a Conditional Gaussian Bayesian Network (BN) model for Dark Triad, aggression, and empathy and to examine the validity of the constructed models as predictable. First, this study conducted a web questionnaire survey using the Short Dark Triad, Buss-Perry Aggression Questionnaire, and Multidimensional Empathy Scale. The BN models were automatically constructed obtained from 637 Japanese participants (372 males, 265 females, age range 22–69 years, M = 50.4, SD = 11.6). The model revealed that (1) psychopathy had a direct and positive arc on aggression types (physical aggression, hostility, and anger) and was negatively linked to other-oriented emotional reactivity and perspective-taking among empathy; (2) narcissism negatively predicted hostility and had a direct positive association with verbal aggression; and (3) Machiavellianism indirectly associated with aggression. Furthermore, the leave-one-out cross-validation (LOOCV) simulation revealed that the performance metrics (accuracy, recall, precision, and F1 score) for predicting the three Dark Triad traits of the test set were approximately 60 % to 70 % (chance level 50 % for High/Low). The precision reached a maximum of 80.4 %. This study revealed that BNs are a powerful and useful tool through automatic data-driven exploration of relationships between the Dark Triad and other personality traits. •Bayesian Network modeling for Dark Triad, aggression, and empathy•Psychopathy was linked positively and negatively to aggression and empathy, respectively.•Narcissism negatively predicted hostility but positively predicted verbal aggression.•Machiavellianism suggested an indirect relationship with aggression (anger and hostility).