Divergent generative AI pathways in higher education: a parallel mediation analysis of autonomous use and human–machine synergy

Yan Cheng, Xinran Liu, Haibo Liu

Introduction The integration of Generative Artificial Intelligence (GenAI) into higher education has raised important questions regarding its influence on student learning outcomes. Drawing on Self-Regulated Learning (SRL) theory, this study examines how learning motivation (LM) relates to knowledge application ability (KAA) through two parallel pathways: human-machine synergy (HMS) and autonomous learning technology (ALT) use. Methods Data were collected from 760 Chinese university students and analyzed using structural equation modeling with bias-corrected bootstrapping. Results The results indicate that LM has a significant positive direct effect on KAA. Among the two mediating pathways, ALT use demonstrates a small but significant partial mediating effect, whereas the HMS pathway is not statistically significant. Discussion These findings suggest that motivated learners may benefit from using GenAI as a self-directed learning support tool, while deeper forms of human-AI collaboration do not necessarily generate stronger knowledge application outcomes. A possible explanation is that excessive cognitive offloading and insufficient algorithmic literacy may weaken the effectiveness of synergistic AI interaction. This study contributes to the literature by highlighting the conditional nature of AI-supported learning and emphasizing the importance of maintaining cognitive autonomy in AI-assisted educational environments.