EMPIRICAL ASSESSMENT OF METACOGNITIVE OUTCOMES: EVALUATING THE IMPACT OF AI-DRIVEN SCAFFOLDING ON UNIVERSITY STUDENTS' LEARNING PROCESSES

students frequently demonstrate significant difficulties in deploying these skills within digital environments. This study examines whether AI-driven scaffolding, configured as cognitive friction, can optimize metacognitive awareness by stimulating critical reflection rather than algorithmic passivity. Through a quasi-experimental test-retest design involving 79 Primary Education students, participants designed authentic assessment instruments based on continuous critical validation of AI outputs against pedagogical goals, thereby transforming evaluation into a genuine metacognitive laboratory. MAI pre/post-test results show that 83.9% perceived AI as a reflective stimulus, while Information Management and Planning capabilities improved significantly. Meta-analytic evidence confirms that GenAI alone does not automatically improve metacognition (g = 0.078), whereas structured scaffolding produces large effects (g = 1.104). These findings confirm that cognitive friction within authentic evaluative contexts transforms human-AI interaction into metacognitive growth, equipping future teachers for conscious instructional design.