Development and Implementation of a Custom GPT Model for Assessing Critical Thinking Skills in Higher Education

This study developed and evaluated a Custom GPT model designed specifically to assess the critical-thinking skills of higher-education students, focusing on synthesis (C4), evaluation (C5), and creation (C6) based on Bloom’s Taxonomy. Employing a design-based research (DBR) methodology, six iterative design cycles were systematically conducted, involving direct classroom observations, expert validations, and practical testing with fourteen mathematics education students. Findings demonstrated that the Custom GPT system is highly practical regarding ease of use (86.7%) and efficiency (85.1%), and practically engaging for learners (81.8%). Quantitative analyses showed significant improvements in students’ critical-thinking skills, particularly within synthesis and evaluation domains, although variations in creation domain performance indicated the need for incorporating more real-world, application-based scenarios. Qualitative feedback confirmed that real-time AI-driven feedback significantly enhanced students' self-reflection and learning strategies. This study contributes theoretically by integrating AI with Bloom’s Taxonomy-based assessment and offers practical implications by providing a scalable, objective, and innovative assessment approach for enhancing higher-order thinking in higher education contexts.