<![CDATA[The emergence of Generative Artificial Intelligence (GenAI) in chemistry education presents both opportunities and challenges for learners' independent thinking and metacognitive awareness. This study aims to analyze in depth the dynamics of students' metacognitive strategies encompassing Planning (P), Monitoring (M), and Evaluating (E) domains and the reconstruction process of stoichiometry concept understanding through interaction with two distinct AI interaction scenarios: Standard AI (linear explanation) and Socratic AI Devil's Advocate (a four-stage approach: Challenge, Probe, Conflict, and Resolve). A descriptive qualitative design with an embedded quantitative element (embedded quantitative design) was employed. The subjects comprised 60 Biology Education students at Universitas Gunung Leuser Kutacane, equally divided into a Standard AI Group (N = 30) and an AI Devil's Advocate Group (N = 30). Data collection was conducted from July 6 to 23, 2026, using four instruments: reflective Student Worksheets (LKM), digital dialogue transcripts (AI log chats), a stoichiometry diagnostic test (pretest and posttest), and a Metacognitive Awareness Inventory (MAI). Qualitative data were analyzed using thematic coding techniques based on metacognitive indicators, while quantitative data were evaluated using Normalized Gain (N-gain) calculations and metacognitive calibration accuracy. The results indicated that interacting with Standard AI tended to induce an illusion of knowing, wherein students' metacognitive activity stalled at a confirmatory Planning level with a moderate N-gain (0.46). Conversely, the AI Devil's Advocate intervention effectively elicited cognitive dissonance that triggered intensive Monitoring and Evaluating, thereby fostering robust stoichiometry concept remediation with a high N-gain (0.77) and realistic metacognitive calibration accuracy. These findings underscore the critical role of pedagogical prompt engineering through a Socratic approach to mitigate cognitive offloading in higher education STEM learning]]>
Copyrights © 2026
