A mixed-methods study on generative AI’s influence on problem-solving mechanisms among pre-service teachers

This mixed-methods study examines how a generative AI (GAI)-supported intervention reshapes pre-service teachers’ (PSTs) problem-solving. Using an explanatory sequential design, 108 PSTs were assigned to no-technology, search engine, or GAI-supported conditions. Quantitatively, the GAI-supported group demonstrated significantly higher metacognition and self-efficacy and lower cognitive load. Qualitatively, thematic analysis revealed a tripartite mechanism: externalizing metacognitive reflection, offloading cognitive tasks, and fostering confidence partially contingent on the AI partner—termed Externalized Self-Efficacy. These findings suggest that while GAI-supported collaboration offers cognitive-affective benefits, it also risks efficacy-source externalization, underscoring the need for teacher education to cultivate critical human-AI partnerships that safeguard endogenous professional agency.