A Generative AI-Based Instructional Strategy for Critical Thinking and Epistemic Autonomy: A Pretest–posttest Study

This study examined a generative artificial intelligence (GenAI)-based instructional strategy using pretest and posttest responses from 96 undergraduate students, with 48 in each condition. The eight-week program emphasized independent verification, argument critique, and reflective justification. Outcomes were a 20-item critical-thinking score (0–40) and an 18-item epistemic-autonomy mean (1–5). Both pretests were controlled in the posttest models. The joint group contrast yielded Wilks lambda = .674, F(2, 91) = 21.99, p < .001. Adjusted differences were 6.82 critical-thinking points (95% CI [4.17, 9.47]; partial eta squared = .221) and 0.79 autonomy points (95% CI [0.53, 1.04]; partial eta squared = .295). Both outcome tests remained significant after Holm adjustment, and HC3 sensitivity intervals excluded zero. Students in the GenAI condition had higher adjusted posttest scores on both outcomes. The findings support further investigation of verification-centered GenAI instruction, with interpretation limited by the study design, measurement evidence, and the absence of reported delayed transfer outcomes.