<![CDATA[Moral degradation poses significant challenges across social, organisational, and digital environments, yet empirical tools for predicting individual vulnerability to unethical behaviour remain limited. This study develops an interpretable machine learning-based predictive model to estimate tendencies toward moral degradation using multidimensional moral domain scores derived from the Moral Perspectives and Foundations Scale (MPFS), with a specific focus on the Perpetrator Relevance (PR01) block. The final analytical sample consisted of 2,130 respondents after data filtering. Two tree-ensemble algorithms, Random Forest (RF) and Gradient Boosting (GB), were implemented and compared using an 80:20 train-test split. Model performance was evaluated using the coefficient of determination (R²), Mean Absolute Error (MAE), and Mean Squared Error (MSE). The results demonstrate that both models achieved strong predictive performance across all PR01 moral domains, with GB consistently outperforming RF. The highest predictive accuracy was observed in the Loyalty (GB R² = 0.616) and Authority (GB R² = 0.595) domains, accompanied by lower MAE and MSE values, indicating stable predictive tendencies rather than deterministic moral behaviour. To enhance interpretability, SHAP analysis was applied, revealing that binding moral dimensions, particularly Loyalty and Authority across multiple moral perspectives, exert the strongest influence on predicted moral degradation tendencies. Overall, the findings highlight the value of integrating ensemble learning with explainable AI techniques in moral psychology. Given the cross-sectional nature of the data, the proposed framework is intended as a risk-detection tool rather than a diagnostic or causal model, while future research should incorporate longitudinal and behavioural data to strengthen generalisability and inference.]]>
