The influence of institutional policies and ChatGPT usage levels on dependence through utilization in higher education

Abstract

The rapid adoption of generative artificial intelligence (AI), particularly ChatGPT, has transformed learning practices in higher education while raising concerns about student dependency and academic integrity. Although previous studies have primarily examined the effects of AI usage intensity, limited attention has been given to how institutional policies shape AI utilization and subsequently influence dependency. This study addresses this gap by examining the effects of ChatGPT usage level and institutional policies on student dependency, with ChatGPT utilization as a mediating variable. A quantitative approach was employed involving 113 students from Politeknik Ilmu Pelayaran Semarang, Indonesia, selected through stratified random sampling. Data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) with SmartPLS 3. The findings indicate that institutional policies significantly enhance ChatGPT utilization but do not directly affect dependency. In contrast, ChatGPT utilization has a significant negative effect on dependency, suggesting that higher-quality utilization reduces the likelihood of excessive reliance. The level of ChatGPT usage does not significantly influence either utilization or dependency. Furthermore, ChatGPT utilization fully mediates the relationship between institutional policies and dependency. These findings extend the literature on AI adoption and technology dependency by demonstrating that the quality of AI utilization, rather than usage intensity, is the primary mechanism through which institutional policies influence dependency. This study suggests a shift from frequency-based perspectives toward utilization-oriented approaches in understanding responsible AI adoption in higher education. Practically, higher education institutions should prioritize policies that promote AI literacy, establish clear ethical guidelines, integrate ChatGPT into learning activities with appropriate supervision, and encourage reflective rather than passive AI use to maximize educational benefits while minimizing dependency.