Effects of generative AI-mediated socratic dialogue on preservice physics teachers’ critical thinking in kinematics

Initial physics teacher education frequently approaches kinematics problem solving as a procedural activity, with limited engagement in conceptual reasoning, argumentation, and metacognitive regulation. This study examines the effects of Socratic dialogue mediated by generative artificial intelligence (GAI), compared with teacher-guided Socratic dialogue alone, on discipline-specific critical thinking (CT) in one-dimensional kinematics. A concurrent mixed-methods, within-group preexperimental design was implemented with 18 preservice physics teachers working in small groups across two sequential phases using parallel kinematics problem sets. Because the instructional conditions were implemented in a fixed order, the observed differences cannot be disentangled from possible practice, learning, and carryover effects. In the first stage, students engaged in instructor-guided Socratic dialogue; in the second stage, the same structure was maintained while incorporating GAI-mediated Socratic support under instructor supervision. Critical thinking was assessed through an analytic rubric applied to written group solutions and complemented by Likert-scale surveys and open-ended responses. Rubric scores showed consistent improvement during the GAI-mediated phase across all groups, with no observed declines. Students also reported higher self-efficacy and intrinsic motivation, while cognitive load exhibited heterogeneous responses associated with the need for verification and epistemic control when interacting with GAI. These results suggest that integrating generative AI into a teacher-mediated Socratic framework can support critical thinking in kinematics, although the sequential preexperimental design requires cautious interpretation.