<![CDATA[Purpose – This study aims to predict and classify the level of student dependency on ChatGPT in completing academic tasks using the Naive Bayes algorithm to support data-driven decision making in higher education.Methods – A quantitative survey approach was employed involving 254 active undergraduate students from the Department of Informatics and Computer Engineering at a public university in Indonesia. Data were collected through a Likert-scale questionnaire measuring five behavioral indicators: purpose of ChatGPT use, interaction frequency and duration, understanding of generated outputs, trust in AI responses, and learning independence. The collected data were cleaned, numerically encoded, and labeled into three dependency categories (low, medium, high). A Naive Bayes classification model was implemented using Orange Data Mining and evaluated under three data split scenarios: 90:10, 80:20, and 70:30.Findings – The results indicate that the 70:30 data split achieved the highest classification performance, with an AUC value of 0.973, accuracy of 85.3%, F1-score of 0.866, and precision of 0.909. These results demonstrate that the Naive Bayes algorithm is effective in identifying distinct patterns of student dependency on ChatGPT based on multidimensional behavioral data.Research limitations – This study is limited to a single academic program and relies on self-reported questionnaire data, which may constrain the generalizability of the findings across different educational contexts.Originality – This study provides empirical evidence on the application of probabilistic classification models to assess student dependency on generative AI, contributing to educational decision sciences by informing institutional policies on balanced and responsible AI use in higher education.]]>
Copyrights © 2026
