From Assistance to Academic Dependency: An Analysis of Students’ Use of ChatGPT in Homework Completion

Abstract

The rapid expansion of generative artificial intelligence has significantly transformed students' learning practices, particularly in the completion of homework assignments. Since the public release of ChatGPT, concerns have emerged regarding its pedagogical value, potential academic dependency, and implications for academic integrity. This study investigates the extent and forms of ChatGPT use among lower and upper secondary school students, focusing on the boundary between academic assistance and dependency. The study also considers the concept of cognitive offloading, referring to students’ reliance on AI tools to reduce cognitive effort during task completion. A mixed-methods explanatory sequential design was employed, combining quantitative data from 132 students (grades VI–XII) with qualitative interview insights. Statistical analyses included descriptive statistics, ANOVA, regression analysis, exploratory factor analysis (EFA), and mediation analysis (model with bootstrapping). Results indicate that ChatGPT is widely used primarily as a supportive tool rather than a full replacement for independent work. Academic dependency levels were generally low; however, regression and mediation analyses revealed that dependency partially mediates the negative relationship between ChatGPT use and independent learning engagement (44% of the total effect). Effect sizes ranged from small to moderate, suggesting a practical but not dominant influence. Findings suggest that ChatGPT itself is not inherently detrimental; rather, its impact depends on the intensity and manner of use. The study highlights the need for structured pedagogical guidance and clear ethical frameworks to ensure responsible integration of generative AI in secondary education.