Analysis and Visualization of Generative AI User Dependency Patterns in Scientific Manuscript Writing

Abstract

The use of Generative Artificial Intelligence (Generative AI) has become an integral part of academic activities, particularly in scientific manuscript writing. Students of the Faculty of Information Technology (FTI) are among the most active users, relying on this technology to enhance productivity and efficiency. However, its integration also brings the risk of dependency, which may affect idea originality and critical thinking. This study examines the influence of Perceived Ease of Use and Perceived Usefulness on AI Dependency, mediated by AI Use Patterns and Attitude Toward Use. It not only identifies generative AI users but also analyzes and visualizes dependency patterns in academic writing. Using a cross-sectional survey and descriptive quantitative analysis, the findings reveal intensive AI usage among FTI students, dominated by Informatics (69%) and Information Systems (31%), with ChatGPT and Gemini as the preferred tools. Structural model analysis with SmartPLS shows a significant positive effect of perceived ease and usefulness on dependency levels (β = 0.704, p < 0.05). Most students (62%) fall into the moderate dependency category, indicating productive yet controlled use. These results highlight the need for ethical literacy and academic awareness to ensure AI supports, rather than diminishes, writing quality and originality.