The Impact of Artificial Intelligence Literacy on Students’ Social-Emotional Competence: An Empirical Study Using Random Forest Feature Importance Modeling

With the rapid diffusion of generative artificial intelligence in basic education, AI literacy is increasingly shaping the social-emotional competence of primary and secondary school students; however, the core drivers underpinning its positive and negative effects remain insufficiently understood. Using 195 valid survey responses, this study aims to construct differentiated predictive models to uncover the mechanisms underlying the dual impacts of AI literacy. Fourteen AI literacy subdimensions were identified and categorized into nine positive and five negative dimensions, supplemented by demographic variables as controls. Two Random Forest classification models were then developed to predict perceived positive and negative impact levels (low/medium/high), and standardized feature-importance scores and heatmaps were used to identify key predictors. The results reveal a complete non-overlap between the driving features of the two impact types: positive impacts are primarily driven by “AI role-playing activities enhancing self-awareness,” “human-AI collaborative projects fostering cooperation,” and “AI ethics education strengthening online responsibility,” whereas negative impacts are dominated by “premature reliance on AI reducing critical thinking,” “generative AI interactions increasing machine preference,” and “excessive AI use leading to emotional dependency.” All demographic variables exhibit importance scores below 0.04, indicating that functional attributes of AI literacy outweigh demographic differences in shaping public perceptions. These findings validate the independence of positive and negative impact mechanisms and provide empirical support for “enhancing strengths while mitigating risks” in AI education, offering actionable insights for promoting students’ social-emotional development.