KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 7 Serial Number 42 2026 02 10 Explainable AI Analysis of Cognitive Distortions and Their Predictive Role in Adolescent Major Depressive Episodes Explainable AI Analysis of Cognitive Distortions and Their Predictive Role in Adolescent Major Depressive Episodes 1 11 10.61838/kman.jayps.5082 EN Journal Article 2025 09 22 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> The present study aimed to investigate the predictive role of cognitive distortions in adolescent major depressive episodes using explainable artificial intelligence techniques to enhance both classification accuracy and interpretability of cognitive risk factors.</p> <p><strong>Methods and Materials: </strong>A cross-sectional predictive-correlational design was employed with a sample of 612 adolescents aged 13 to 18 years recruited from secondary schools in Georgia through multistage cluster sampling. Cognitive distortions were assessed using a validated self-report inventory measuring catastrophizing, overgeneralization, personalization, mind reading, and dichotomous thinking. Major depressive episodes were identified using a structured screening protocol based on DSM-5 criteria supplemented by the PHQ-9 adolescent version. Data analysis integrated traditional statistical methods and supervised machine learning algorithms. The dataset was divided into training and testing subsets using stratified sampling. Logistic regression, support vector machine, random forest, multilayer perceptron, and gradient boosting (XGBoost) models were implemented with cross-validation and hyperparameter tuning. Model performance was evaluated using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). Explainability was achieved using SHAP (Shapley Additive Explanations) to determine feature importance and nonlinear effects.</p> <p><strong>Findings: </strong>Cognitive distortions were significantly and positively associated with depressive symptoms (p &lt; 0.01). Machine learning models demonstrated high predictive accuracy, with the XGBoost model achieving the strongest performance (AUC = 0.95). SHAP analysis revealed that catastrophizing, overgeneralization, and mind reading contributed the highest predictive weight to classification outcomes. Nonlinear threshold effects indicated substantially increased depression probability beyond upper-quartile distortion scores.</p> <p><strong>Conclusion: </strong>Cognitive distortions represent powerful and interpretable predictors of adolescent major depressive episodes, and the integration of explainable</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/jayps/article/download/5082/9161