Mapping University Students Risk of Dependence on Generative Artificial Intelligence Using K-Means Clustering

Abstract

The growing use of generative artificial intelligence (GenAI) in higher education has changed how university students approach academic tasks. Although these tools can support learning and improve efficiency, frequent reliance on them raises concerns about students' ability to study and solve problems independently. This study aims to identify patterns of GenAI use among university students and explore potential indicators of dependence through K-Means clustering. An exploratory quantitative approach was adopted, using questionnaire responses collected through 15 items measured on a five-point Likert scale. Of the 106 responses received, 104 met the inclusion criteria and were retained for analysis. The data underwent screening, numerical encoding, and reverse scoring before being grouped using the K-Means algorithm. The appropriate number of clusters was selected by comparing Silhouette Scores across solutions containing two to seven clusters. The analysis identified three student groups, with a Silhouette Score of 0.2443. The first group comprised 30 students (28.85%) who showed relatively higher GenAI use and stronger indications of behavioral dependence. The second group included 11 students (10.58%) who reported lower GenAI use but demonstrated limitations in academic independence and understanding. The remaining 63 students (60.58%) formed a group characterized by moderate GenAI use. These findings suggest that frequency of use alone does not adequately reflect the risk of dependence. However, the relatively low Silhouette Score indicates limited separation between groups, meaning that the resulting profiles should be interpreted as exploratory rather than definitive.