From AI Use to Critical Thinking Among Medical Students: A Moderated Mediation Perspective on Cognitive Load and Self-Regulated Learning

Synthesis: Arshad et al. (2026) examined how AI-based educational technology influences critical thinking among 480 undergraduate medical students in Pakistan, using a cross-sectional design and Hayes' PROCESS Model 14. They found that AI use was positively associated with critical thinking and self-regulated learning, while cognitive load negatively related to both. Cognitive load partially mediated the AI-use→critical-thinking link, and self-regulated learning significantly moderated that indirect effect — the negative impact of cognitive load on critical thinking weakened at higher levels of self-AI Regulation in Education. The study argues that AI's effectiveness for higher-order thinking depends not only on the cognitive support AI provides but on learners' capacity to regulate their engagement.

Key Findings

• AI-based technology use was positively associated with both critical thinking and self-regulated learning among medical students, whereas perceived cognitive load was negatively associated with both outcomes.
• Mediation analysis showed cognitive load partially mediated the relationship between AI use and critical thinking — part of AI's effect on thinking runs through the cognitive burden it imposes.
• Moderated mediation revealed self-regulated learning significantly moderated the indirect effect: the negative impact of cognitive load on critical thinking was weakened at higher levels of self-regulation.
• The study situates AI's educational impact within Cognitive Load Theory and Self-Regulated Learning, framing generative AI tools (ChatGPT, Claude, Gemini) as double-edged — capable of reducing extraneous load by organizing/summarizing, yet adding burden through large volumes of content that must be evaluated and validated.

Study Design & Method

The authors used a cross-sectional survey design with 480 undergraduate medical students who had prior experience using AI tools. Data were collected with standardized measures assessing AI usage, cognitive load, self-regulated learning, and critical thinking. Hayes' PROCESS Model 14 was used to test mediation and moderated mediation effects. The analytic approach allowed the authors to model cognitive load as a mediator of AI use on critical thinking and self-regulated learning as a moderator of the mediated (indirect) path — specifically, whether the load→thinking pathway is conditional on students' regulatory capacity.

What this means for practice

• Instructors. Design AI-supported tasks so students must verify, reflect on, and judge what the tool produces rather than accepting answers wholesale; the mediation analysis shows part of AI's effect on Critical Thinking runs through the cognitive load it imposes.
• Instructors. Teach planning, monitoring, and evaluation of AI-assisted work explicitly. Self-Regulated Learning was the moderator that weakened the negative load→thinking pathway, so regulatory skill is the lever students can actually practice.
• Faculty developers. Audit AI activities for how much mental work they remove. Summarizing and organizing can cut extraneous load, but large volumes of AI-generated content add evaluation and validation burden of their own.
• Administrators. Do not adopt AI tools on efficiency, performance, or engagement metrics alone. This study modeled the cognitive processes underneath outcomes and connects AI use to academic integrity and dependency concerns in medical education.
• Researchers. Treat the field's emphasis on efficiency and short-term performance as a gap to correct; measure the higher-order thinking and load processes through which AI effects travel.

Limitations

• The cross-sectional, non-experimental design limits causal inference despite theory-consistent directional modeling.
• Data are self-report measures collected at a single time point, which raises common-method-bias concerns, and the moderated mediation effects are estimated from that single snapshot.
• The sample is drawn from medical students in one national context (Pakistan), bounding generalizability to other disciplines, institutions and educational systems.
• AI usage, cognitive load and self-regulated learning are all operationalized via standardized self-report instruments rather than objective behavioral measures.

Citation

Arshad, A., Lone, A., Arickswamy, L., Hassan, K., Alnaim, A. A., & AlFarhan, M. F. (2026). From AI use to critical thinking among medical students: A moderated mediation perspective on cognitive load and self-regulated learning.