Human-Centered AI Metacognition: Development and Validation of a Scale for Preserving Cognitive Autonomy in Human-AI...

The rapid advancement of generative artificial intelligence (AI) is transforming human cognitive activities and creating a growing need to preserve cognitive autonomy in human-AI interactions. Although recent research has examined metacognition in AI-mediated environments, existing work has primarily focused on improving AI use and human-AI collaboration, with less attention to how individuals regulate the role and extent of AI involvement in their own cognitive activities. To address this gap, the present study introduced Human-Centered AI Metacognition (HC-AIM), defined as the capacity to recognize AI capabilities and limitations, monitor AI-generated information, and regulate the extent to which AI participates in one's own cognitive activities while maintaining cognitive autonomy. Across three studies, we developed the Chinese version of the Human-Centered AI Metacognition Scale (HC-AIMS) and conducted an initial psychometric evaluation. Study 1 examined the scale structure using exploratory factor analysis and supported a predominantly unidimensional structure with good internal consistency. Study 2 examined this structure using confirmatory factor analysis and provided evidence of its convergent validity through its association with general metacognition. Study 3 further replicated the scale structure and internal consistency and examined its relationships with AI literacy, collaborative AI metacognition, and autonomous cognitive offloading. HC-AIM was positively associated with all three constructs. Comparisons of theoretically specified factor models provided preliminary evidence of empirical distinguishability between HC-AIM and both AI literacy and collaborative AI metacognition, despite substantial correlations with these constructs. HC-AIM was also positively associated with autonomous cognitive offloading, providing complementary self-reported criterion-related evidence for its validity. Taken together, the findings suggest that HC-AIM represents a contextualized, human-centered form of metacognitive regulation that is closely related to broader metacognitive and AI-related competencies while specifically emphasizing the regulation of AI participation in human cognition. The Chinese version of the HC-AIMS provides a theoretically grounded starting point for investigating human cognition and cognitive autonomy in AI-mediated environments.