Understanding ChatGPT dependency in Science student learning: a structural model approach

ChatGPT, an AI-based tool for generating text responses, has become popular among university students to aid in understanding complex subjects. This study explores factors influencing ChatGPT adoption among Indonesian science students, focusing on ease of use (EU), perceived usefulness (PU), hedonic motivation (HM), satisfaction of use (SU), and dependency of use (DU). A quantitative method was used to collect data from 205 students across 22 provinces via convenience sampling, analyzed using PLS-SEM. Results show that the EU significantly affects SU but not DU. PU and HM significantly influence SU and DU, indicating their key roles. However, SU does not significantly affect DU, suggesting that satisfaction alone does not drive dependency. These findings highlight the importance of balanced ChatGPT integration in education to support learning without encouraging over-reliance. This study offers empirical insights from a developing country for educators and policymakers on responsible AI use in science education.