Read the preprint on EdArXiv (open access, CC BY 4.0)
By Genevive Bjorn, Ed.D. (ORCID 0000-0002-2856-1067), Higher Learning Lab
Citation: Bjorn, G. (2026). Artificial co-regulation: AI tutors that never fade capture learners’ regulation instead of transferring it [Preprint]. EdArXiv. https://doi.org/10.35542/osf.io/qgv4t
Status: Preprint posted 31 October 2026. Submitted to Educational Psychology (Taylor & Francis), special issue “AI in Educational Psychology”; under review.
About this paper
Conversational AI tutors are now used by most university students and are being deployed across whole school systems, while field experiments show students performing better with the tutor present and worse without it. This position paper names the mechanism: artificial co-regulation. Human scaffolding works because it fades. The tutor steps back as the learner grows, and attachment theory explains why learners let go. Conversational agents recruit the learner’s attachment system but are neither designed nor incentivized to fade, so regulation migrates to the agent instead of transferring to the learner.
The paper traces the cost across reading development, from the fourth-grade transition to disciplinary reading at university, with multilingual and first-generation students as the central case. It states what would refute the model and proposes five design and measurement standards, contingent fading first among them, before “permanent AI partner” models are adopted for learners who are still becoming capable of partnership.
Keywords: generative AI; scaffolding; attachment; co-regulation; epistemic agency; academic literacy
Related resource: Artificial Intelligence, Learning, and Attachment in K–12 Education: A Review of the Research, the Lab’s practitioner review of the evidence this paper builds on.
New research reviews
Get notified when a new review or free resource is published
A handful of emails a year for educators, counselors, and grant writers. No spam, and you can leave any time.
