Polished Artifacts, Fragile Engagement? Tackling the Challenge of Reduced Epistemic Effort in Human-AI Knowledge Construction

Learning and knowledge construction are socially organized processes of meaning-making that rely on co-regulation and epistemic agency. Over the past decades, research in Computer-Supported Collaborative Learning (CSCL) has developed a rich theoretical repertoire to explain how these processes emerge through mediated interaction. This article revisits these foundational assumptions against the background of people’s use of generative AI (genAI). As epistemic processes are increasingly supported by genAI, we realize a risk of superficial understanding and the illusion of comprehension. While genAI can efficiently produce and refine knowledge-related artifacts, it may simultaneously reduce human epistemic effort by shifting regulatory and evaluative processes toward the AI. Drawing on CSCL theories and research traditions, we conceptualize this risk through two different strands: a social-cognitive and an artifact-oriented account. The social-cognitive strand is grounded in automation bias, that is, people’s systematic tendency to attribute greater epistemic competence to AI systems than to themselves. This shift may weaken shared regulation, transactivity, and epistemic agency. The artifact-oriented strand focuses on the epistemic properties of polished external artifacts that resemble finalized products rather than provisional drafts. Such artifacts may induce a sense of epistemic closure, reducing cognitive conflicts and deeper elaboration. We describe these intricacies as key challenges for the next phase of CSCL research and appeal to structure AI participation in ways that preserve conflict and iterative refinement by conceptualizing AI as an argumentative partner or epistemic challenger to enhance epistemic quality without diminishing human epistemic effort.