Unconstraining Metacognition in Human-AI interaction: MET-AI

AI systems have the potential to significantly enhance human cognition, yet their effectiveness is often constrained by a lack of support for metacognition, the ability to plan, monitor, and regulate one's own thinking. My research has shown that while AI tools can improve performance, they frequently impair users' ability to evaluate their decisions, leading to overconfidence and diminished agency. The MET-AI project aims to address this challenge by exploring how metacognition can be optimized in human-AI interaction. Through experimental studies, computational modeling, and design interventions, I will investigate how users monitor and control their thinking when working with AI. Building on pilot studies that reveal AI-induced biases in metacognitive processes, the project will develop theoretical models and practical guidelines for adaptive AI interfaces that empower users to think critically and perform optimally.