Agentic AI in Education: Artificial Agency and the Social-Cognitive Dynamics of Learning

Editors

• Lixiang Yan, Tsinghua University
• Yizhou Fan, Peking University
• Yueqiao Jin, Monash University
• Nancy Law, University of Hong Kong
• Yu Zhang, Tsinghua University
• Xibin Han, Tsinghua University

Description

This special section examines agentic AI in education, defined as computational systems exhibiting interactivity, bounded autonomy, and adaptability within human-defined objectives. Unlike reactive AI tutoring systems or prompt-based generative AI tools, agentic AI initiates action, sustains participation, and maintains functional roles through persistent memory and task decomposition. The shift toward such systems reconfigures how epistemic labour and coordination are distributed within learning communities, affecting learners' perceptions of agency, shared regulation, and collaborative reasoning.

The section welcomes empirical and conceptually grounded studies exploring how agentic AI reshapes learner perceptions, teacher roles, epistemic labour, and social-cognitive dynamics. Particular emphasis is placed on understanding when artificial agency supports productive learning versus constraining learner autonomy, examining this across diverse educational contexts, learner populations, and cultural settings. Contributions addressing governance, responsibility, equity, and methodological innovation are encouraged.

Extrait. Lire l'appel complet sur le site de l'éditeur.

Potential topics

• Conceptualising Agentic AI as Artificial Agency in Education: Frameworks and models that define agentic AI explicitly in terms of interactivity, bounded autonomy, and adaptability, and that clarify boundaries between agentic AI, generative AI assistants, and reactive AIED systems.
• Learner and Teacher Perceptions of Artificial Agency: Empirical investigations of how learners and educators interpret AI initiative, persistence, and role adoption, and how these perceptions shape trust, attribution of intent, responsibility, and participation in learning activities.
• Epistemic Labour and Initiative Redistribution: Studies examining how agentic AI redistributes epistemic labour within individual or collaborative learning, including effects on problem framing, task coordination, evaluation, and decision-making when AI initiates actions rather than responding to prompts.
• Regulation, Coordination, and Social-Cognitive Dynamics: Analyses of how agentic AI influences self-regulation, co-regulation, and shared regulation, particularly in group settings where AI adopts persistent roles that shape interactional norms, turn-taking, and closure of inquiry.
• Temporal Persistence and Path Dependence in Learning Processes: Research that captures how agentic AI's sustained participation and memory create path-dependent learning trajectories, including benefits for coherence and risks of early framing effects or cascading errors.
• Identity Inference, Social Presence, and Responsibility: Empirical work on how proactive AI behaviour affects identity inference and social presence in cue-lean environments, and how responsibility and accountability are negotiated when learning actions originate from artificial agents.
• Methodological Approaches for Studying Artificial Agency: Methodological contributions that advance the measurement of artificial agency and its effects, including multimodal analyses, process-level modelling, and designs that distinguish agentic participation from tool-based assistance.
• Governance and Educational Implications of Agentic AI: Analyses addressing accountability, transparency, and institutional responsibility when agentic AI functions as a collaborator, coordinator, or evaluator in educational settings.