Identification of Key Factors Influencing Teachers’ Self-Perceived AI Literacy: An XGBoost and SHAP-Based Approach

The rapid advancement of digital technologies and artificial intelligence (AI) is reshaping K-12 education, thereby emphasizing the growing need for AI Literacy among teachers. This study identifies key factors that influence teachers’ self-perceived AI Literacy and evaluates their impact on AI Literacy performance across various teaching phases, using Extreme Gradient Boosting (XGBoost) and Shapley Additive Explanations (SHAP). Data collected from 1172 K-12 teachers in South Korea were preprocessed and then split into an 80:20 training-to-testing ratio. To optimize model performance, Bayesian Optimization was used to fine-tune key hyperparameters, including the learning rate, maximum depth, subsample ratio, and number of boosting rounds. The model’s predictive accuracy was assessed using R2, MSE, MAE, and RMSE. The optimized model achieved R2 values of 0.8206 (Class Preparation), 0.8007 (Class Implementation), 0.8066 (Class Assessment), and 0.7746 (Utilizing Assessment Results). The results indicate that technical knowledge and AI Literacy skills are the most influential factors in the Class Preparation and Implementation Phases, while educational decision-making and ethical considerations play a crucial role in the Assessment and Utilizing Assessment Results Phases. Further, SHAP analysis highlights that both teachers’ and students’ perceived levels of AI learning significantly impact the adoption of AI Literacy, underscoring the importance of contextual factors in integrating AI within education. These findings emphasize the need for AI Literacy education that integrates technical competencies, pedagogical strategies, and ethical decision-making. This study provides empirical insights to support the development of teacher training programs and AI Literacy policies, ensuring the effective integration of AI in education.