This study investigates the prevalence of Generative Artificial Intelligence (AI) usage and its association with critical thinking among medical students. Conducted as a cross-sectional study at a government medical college in Tamil Nadu, India, it found that 96.3% of 722 undergraduate medical students used Generative AI. The study revealed that higher AI dependence, particularly among male and final-year students, was significantly and inversely correlated with lower critical thinking scores. Female students generally demonstrated higher critical thinking. The findings suggest that while Generative AI is widely adopted, excessive reliance may diminish critical thinking, prompting a need for medical curricula to incorporate active learning and clinical reasoning assessments, promoting AI as an adjunct rather than a replacement for human cognition.
Introduction
Generative AI tools like ChatGPT and Google Gemini have revolutionized education but introduce the risk of 'AI dependency,' an excessive reliance on AI that can lead to 'cognitive offloading' and 'cognitive inertia,' diminishing critical thinking. Despite growing evidence of increased AI dependency and its negative impact on critical thinking in general student populations, there's a significant gap regarding medical students, especially in India. This study aims to assess Generative AI dependence and its association with critical thinking among undergraduate medical students, given their demanding curriculum and the implications for future diagnostic reasoning and patient care.
Materials and methods
A cross-sectional observational study was conducted at Tirunelveli Medical College in Tamil Nadu, India, from April to May 2026. The study included undergraduate medical students who provided informed consent. A calculated minimum sample size of 282 was exceeded by the 722 participants (72.2% response rate), with 695 responses used for inferential analysis after excluding non-AI users. Data were collected using a structured digital questionnaire comprising the 11-item Generative AI Dependency Scale (assessing Cognitive Preoccupation, Negative Consequences, and Withdrawal) and the 20-item Critical Thinking Disposition Assessment Questionnaire (CTDAQ-20) (assessing Confidence and Systematicity, Analyticity and Flexibility, and Maturity). Statistical analysis included descriptive statistics, non-parametric tests (Mann-Whitney U, Kruskal-Wallis, Bonferroni correction), Spearman's rank correlation, and multiple linear regression.
Results
The study found a near-universal prevalence of Generative AI usage (96.3%, 695 participants) among medical students, primarily for academic purposes like concept clarification and writing support, with a median usage duration of 12 months. The overall median AI dependence score was 2.1, with male students showing significantly higher dependence than females (p=0.019) and final-year students reporting higher dependence than other years (p<0.001), albeit with small effect sizes. The overall median critical thinking score was 71.0; female students scored significantly higher than males (p<0.001), and first-year students showed better critical thinking than other years (p=0.031). A significant inverse correlation existed between AI dependence and overall critical thinking (rho=-0.248, p<0.001). Multiple linear regression confirmed that higher AI dependence (B=-1.487, p=0.001) and increasing age (B=-0.868, p=0.008) independently predicted lower critical thinking, while female gender (B=2.794, p<0.001) predicted higher critical thinking.
Discussion
The study highlights the widespread use of Generative AI among medical students for academic tasks and a significant inverse relationship between AI dependency and critical thinking, supporting concerns about 'cognitive offloading' and 'cognitive inertia.' Male and final-year students exhibited higher AI dependence and lower critical thinking, possibly due to academic pressures. While statistically significant, the modest effect sizes suggest that critical thinking is influenced by multiple factors beyond AI dependence. The cross-sectional design precludes establishing causality. The discussion advocates for integrating active learning and clinical reasoning assessments like OSCEs and DOPS into medical curricula and promoting AI literacy to encourage responsible AI use as an adjunct rather than a replacement for independent thought, rather than outright prohibition.
Conclusions
Generative AI is almost universally used by undergraduate medical students, and increased dependence on it is independently associated with lower critical thinking abilities, although with modest effect sizes. Female students generally exhibited stronger critical thinking dispositions. The study's cross-sectional design limits causal inferences, suggesting that AI dependence is one of several factors influencing critical thinking. The findings emphasize the need for medical education to evolve by prioritizing clinical reasoning and independent judgment. Further longitudinal, multicenter studies are recommended to fully understand the long-term relationship between AI dependence and critical thinking in medical training, incorporating a broader range of educational, behavioral, and institutional factors.
