Fostering Critical Thinking by Integrating Case-Based Learning and Generative AI in a Closed-Loop Reflective Learning Environment
Abstract
Critical thinking is a core competency in higher education, particularly in learning contexts involving ill-structured problems, uncertainty, and reasoned decision-making. Although case-based learning can foster higher-order thinking, it often scales poorly and provides limited support for timely, structured assessment of students' reasoning. Large language models (LLMs) offer opportunities for scalable learning support, yet many existing AI-based systems lack pedagogically grounded mechanisms for developing critical thinking. To address these gaps, this study proposes an AI-driven closed-loop learning framework that integrates case-based learning with two specialized LLM agents, a Content Generation Agent (CGA) that produces context-rich scenarios and a Feedback Generation Agent (FGA) that evaluates student reasoning and delivers structured, dimension-specific feedback. A mixed-methods study with 38 postgraduate financial planning students found a statistically significant improvement in domain knowledge and strong performance across critical thinking dimensions, especially evaluation and analysis. Students also reported high engagement and perceived learning effectiveness; however, variability in explanation skills and concerns about feedback timeliness highlight areas for improvement. Overall, the findings suggest that AIsupported closed-loop learning is a promising, scalable approach to developing critical thinking in higher education. Index Terms-Critical thinking, Case-based learning, Generative AI, Reflective feedback, Personal finance education I. I NTRODUCTION In higher education, critical thinking develops not through passive knowledge reception, but through repeated cycles of articulating, scrutinizing, and revising one's reasoning [1], [2]. From this perspective, learning is a reflective cognitive process in which learners externalize judgments, receive responses to their thinking, and refine their understanding accordingly [3]. This process, conceptualized as a form of "cognitive echo", is especially important when students confront complex, illstructured problems with no single predetermined solution [4]. In such contexts, the educational challen