From Tool to Crutch: The Longitudinal Impact of AI Dependence on Academic Innovation in Doctoral Training

• In Event: The Use of Advanced Statistical Methods in the Study of Students and the Teaching Workforce Over Time

Abstract

This three-wave longitudinal study (March 2024–January 2025) surveyed 356 PhD students from nine Chinese universities to examine how AI dependence influences academic self-efficacy and innovation. AI dependence was conceptualized across behavioral, emotional, and functional dimensions, while innovation involved the intention and execution of novel research ideas. Cross-lagged panel analysis showed that higher AI dependence predicted lower self-efficacy (β = –.17 to –.24) and reduced innovation (β = –.25 to –.27) over time. Self-efficacy also predicted innovation (β = .10 to .16) and partially mediated the negative impact of AI dependence (indirect β = –.03, p < .01). These findings suggest that over-reliance on AI may undermine doctoral students’ creativity, highlighting the need to foster AI literacy and self-directed academic skills.

Authors

• Yutong Liu, Southeast University

• Jiangtao Han, Xihua University

• Entao Ma, Guangzhou University