Computer self-efficacy and perceived knowledge application in generative AI-supported learning: a cross-sectional parallel mediation analysis

Abstract

Background This study examined whether human–AI synergy (HAS) and autonomous learning technology use (ALT) mediated the association between computer self-efficacy (CSE) and perceived knowledge application ability (KAA). Methods A cross-sectional survey of 801 undergraduates from five higher education institutions in Shandong Province, China, was analyzed using confirmatory factor analysis, structural equation modeling, and bootstrapping. Results CSE was positively associated with HAS and ALT. ALT showed a positive indirect association with perceived KAA, whereas HAS showed a negative conditional indirect association. The direct CSE–KAA association was not significant after including the mediators. Conclusion CSE was associated with perceived knowledge application through two opposite indirect pathways. The negative HAS coefficient should be interpreted cautiously because HAS measured collaboration rather than cognitive offloading, and KAA was self-reported.