AI tutors have emerged as a critical technological intervention to meet the escalating demands for personalized and accessible STEM education. Guided by the Stimulus–Organism–Response (S-O-R) framework, this study investigates the influence of AI tutors on high school students’ learning processes by examining how six key stimuli, including control, efficiency, enjoyment, learning content, personalization, and security, affect the organismic states of learning motivation and satisfaction, which subsequently influence learning outcomes. Using Partial Least Squares Structural Equation Modeling (PLS-SEM) on data collected from 478 Vietnamese high school students, the results reveal differentiated effects across system dimensions. Control operates as a hygiene factor, defined as a baseline functional requirement that prevents dissatisfaction but does not inherently drive engagement, which enhances learning satisfaction but does not significantly predict motivation. In contrast, efficiency, enjoyment, learning content, personalization, and security serve as dual drivers that positively influence both motivation and satisfaction. Critically, both learning motivation and satisfaction significantly contribute to improved learning outcomes. Theoretically, the study advances AI-in-education research by validating a multidimensional AI tutor framework tailored to the K–12 STEM context. Practically, the findings highlight that while hygiene factors like intuitive interfaces and reliable performance reduce user frustration, meaningful engagement depends on adaptive learning pathways, emotionally supportive interactions, and robust data protection. These insights provide actionable guidance for developers and educators seeking to enhance the pedagogical effectiveness and responsible deployment of AI tutors in high school STEM education.
Pham et al. (Mon,) studied this question.