GENERATIVE AI DEPENDENCY AND ACADEMIC PERFORMANCE OF UNIVERSITY STUDENTS: THE MEDIATING ROLE OF CRITICAL THINKING

Abstract

Generative artificial intelligence (AI) tools such as ChatGPT have spread rapidly through Pakistani universities. Many students now use them every day, which raises the question of whether dependence on these tools affects how well students think and perform. This study examined the effect of generative AI dependency on the academic performance of university students in District Lahore, and whether critical thinking mediates this effect. A quantitative, survey research design was used. Through multistage stratified random sampling, 384 undergraduate students were selected from four public and four private universities. Data were collected with a 12-item Generative AI Dependency Scale adapted from Morales-García et al. (2024), a 14-item Critical Thinking Scale adapted from the Critical Thinking Disposition Scale of Sosu (2013), and self-reported cumulative grade point average (CGPA). Both scales showed high reliability (α = .920 and .916). Most students (60.9%) showed moderate AI dependency and 21.1% showed high dependency. Students with high dependency had significantly lower critical thinking (M = 2.57) and CGPA (M = 2.87) than students with low dependency (M = 3.40 and 3.19). AI dependency negatively predicted CGPA (B = −0.138, p < .001) and critical thinking (B = −0.378, p < .001), while critical thinking positively predicted CGPA (B = 0.283, p < .001). Critical thinking fully mediated the effect of AI dependency on CGPA (indirect effect = −0.107, 95% bootstrap CI [−0.145, −0.075]), explaining 77.6% of the total effect. Dependency increased with daily hours of AI use, but no significant differences were found by gender or university sector. The study concludes that AI dependency appears to lower academic performance mainly by weakening critical thinking, and recommends AI literacy training and assessment that requires students' own reasoning.