Family-school autonomy support for children's responsible use of generative artificial intelligence

Synthesis: This mini review argues that the question of children's generative AI use is the wrong frame, and that the productive question is whether the adults around them support autonomy rather than control. Organizing a scattered literature through Self-Determination Theory, the authors distinguish dependent from autonomous Cognitive Offloading, map the psychological pathways from AI use to learning engagement and to Academic Integrity, and set out a developmental research agenda. Their central analytical move is to treat the family-school coordination that most guidance assumes as an untested hypothesis: two literatures on non-overlapping samples have never been tested together, and the review specifies three competing models — additive, synergistic and compensatory — that would discriminate between them.

Key Findings

1. GenAI use is multidimensional rather than a single behavior, and the review proposes that its learning consequences depend on whether offloading is dependent (core thinking delegated) or autonomous (AI scaffolds while the learner keeps epistemic agency). A three-wave study of 589 university students and early-career knowledge workers found dependent offloading associated with transferred agency, lower intrinsic motivation and poorer perceived cognitive outcomes, while autonomous offloading showed the opposite pattern — with immediate performance benefits identical in both cases, making maladaptive use hard to detect in the moment.
2. The evidence on learning outcomes is genuinely split, and the review treats both sides as real: a three-level meta-analysis reported a moderate overall benefit (g = 0.499, and g = 0.669 for comprehension, cognition and Creativity), and a meta-analysis of foreign-language emotions reported a large improvement (g = 0.788, especially for reduced anxiety and boredom); against that sit associations between dependence, fatigue and weaker critical thinking, cases of deep use coexisting with surrendered agency, and a subgroup of English-as-a-foreign-language students relying heavily on generated text.
3. Context is a moderator. Positive associations with engagement were strongest in high-challenge, high-support classrooms, while AI use was negatively associated with engagement in low-challenge, high-support settings — the design of the task, not the tool, carries the sign of the effect.
4. Family and school guidance have been studied on non-overlapping samples. The review formalizes three testable versions of the coordination hypothesis: additive (two independent main effects), synergistic (positive interaction, the version embedded in most current guidance and the one with no supporting evidence) and compensatory (negative interaction with diminishing returns when both settings provide support).
5. Autonomy support is explicitly not permissiveness, and the review flags measurement problems: instruments for AI literacy, dependence and overreliance were largely developed in single-country or specialized university samples, and most studies are cross-sectional and self-reported.

Where the responsibility sits

The synthesised guidance literature pushes risk management onto families and schools while presenting their coordination as established. This review separates what is known from what is assumed. It documents evidence that reflective GenAI use is associated with lower academic impostor syndrome and stronger engagement while unreflective use shows the reverse, that student profiles range from constructive to substitution-oriented, and that academic AI overreliance is better understood as externalisation of cognitive and self-regulatory processes than as technology addiction. It then states plainly that the coordinated family-school model has not been tested and that a factorial trial contrasting family-only, school-only, coordinated and usual-practice guidance would be the strongest test — requiring comparable autonomy-support measures from the same students in both settings, power for an interaction rather than main effects, and behavioral rather than self-reported indices.

Equity and design considerations

The research agenda also names the conditions that make blanket advice unsafe. Comparative work across early childhood, primary school, adolescence and higher education is rare despite developmental theory predicting different risks. Participants in the behavioral-measurement studies are often minors, so privacy-preserving protocols are treated as a precondition rather than an afterthought. The review closes with participatory design and AI Governance as directions, arguing that curriculum work should integrate AI competencies into broader digital literacy while specifying developmental progression.

Connected Concepts

• Self-Determination Theory — the organizing framework for autonomy support
• Cognitive Offloading — dependent versus autonomous offloading is the review's key distinction
• Generative AI — the tool whose use is being regulated
• K-12 — the developmental populations in scope
• Academic Integrity — moral cognition, self-evaluation and policy clarity as pathways
• Motivation — need satisfaction, enjoyment and retained cognitive agency
• Parents and Families

Connected Articles

• Understanding Student Dependency on AI: The Role of AI Literacy, Academic Self-Efficacy, and Resource Management Strategies — Student dependency, AI literacy and self-efficacy
• Meta-Cognitive Insights into Cognitive Offloading: Mechanisms, Interventions, and Educational Implications — Cognitive offloading and metacognition review
• Examining the Impact of Generative AI on Student Motivation and Engagement: The Mediating Role of Autonomy-Support and Autonomous Motivation in Education — Generative AI, motivation and engagement
• Co-designing AI with youth partners: Enabling ideal classroom relationships through a novel AI relational privacy ethical framework — Co-designing AI with youth and relational privacy
• Against frictionless AI — Against frictionless AI

Citation

Fan, Y., Li, X., & Zhang, R. (2026). Family-school autonomy support for children's responsible use of generative artificial intelligence: a self-determination theory synthesis and developmental research agenda. Frontiers in Psychology, 17, 1969217.