AI Sabbatical: A conceptual framework for cognitive sustainability in AI-augmented knowledge work

Abstract

Artificial intelligence (AI) enhances knowledge work through productivity gains and decision support; however, sustained reliance on AI may weaken reflective thinking, independent reasoning, metacognitive monitoring, and epistemic agency through cognitive offloading. Existing responsible AI frameworks largely focus on governing AI systems and optimizing cognition during use rather than deliberately interrupting AI-mediated reasoning. This study introduces the AI Sabbatical to address this gap. This study uses a structured integrative conceptual review to synthesize the literature on generative AI, cognitive offloading, automation bias, metacognition, workplace recovery, and human-centered AI governance. The AI Sabbatical is a structured, temporary, task-sensitive restriction on Generative Artificial Intelligence (GenAI) assistance that preserves independent reasoning and judgment. This study specifies its dimensions and a micro, meso, and organizational typology, distinguishes it from related concepts, conceptualizes cognitive sustainability as a multi-level construct, and develops seven propositions linking AI reliance, cognitive risk, reactivation, and outcomes. Sustainable human–AI collaboration requires governance that preserves human cognitive capability through the structured interruption of AI-mediated reasoning. This study is among the first to conceptualize temporary non-use of AI as a governance mechanism, extending responsible AI governance beyond AI systems to the human capacities and organizational conditions required for sustainable AI-augmented knowledge work.