Self-efficacy, engagement, attitudes, and perceived learning in AI-assisted STEM education

Introduction The rapid advancement of AI technologies is reshaping educational practices, offering new ways to support learning activities. This study examines the interrelationships among self-efficacy, engagement, attitudes, behavioral intentions, and perceived learning in secondary STEM education. Methods The study was based on real-world classroom teaching involving 710 students, who engaged with an AI learning companion that provided personalized guidance and support during STEM laboratory courses. Using a single-group, post-test-only design and partial least squares structural equation modeling (PLS-SEM), the study examined concurrent associations among students' self-reported perceptions. Results Attitude had the largest estimated direct association with perceived learning (β = .306), followed by engagement ( β = .275), intention ( β = .176), and self-efficacy ( β = .119). Discussion Among students who used the AI learning companion, higher reported engagement was associated with higher perceived learning. These findings describe concurrent associations and do not establish that changes in learning are attributable to the AI learning companion.