Identifying Human Dark Triad from Text Data Through Machine Learning Models

Machiavellian, narcissistic, and psychopathological personality characteristics are together referred to as the “Dark Triad†. These characteristics include a lack of empathy, self-centeredness, and manipulative tendencies. Accurately identifying those with these features has significant ramifications for several fields, including psychology, criminology, and human resources. This study investigates the use of machine learning models and natural language processing (NLP) approaches to extract Dark Triad characteristics from text data. To assess the efficacy of several models for detecting people with Dark Triad qualities, we compare their results to ground truth obtained from survey data for Random Forest, K-Nearest Neighbors (KNN), Linear Support Vector Classifier (Linear SVC), Naive Bayes, and Neural Network models. The experiments yield Random Forest, KNN, Linear SVC, and neural network obtained testing accuracies of 83% to 84% for Machiavellianism; KNN achieved 67% for Psychopathy; and Neural Network reached 71% for Narcissism. Overall, the Linear SVC has consistently given better results between 60-80 percent for the three Dark Triad traits, showing constant precision and recall rates. Future aims include refining models, exploring alternative feature extraction methods, and expanding the dataset for improved accuracy and generalizability.