Artificial Intelligence in Toys: Implications for Child Development and Play

Synthesis: Xu, Girouard, and Shi (2026) review how AI-enabled toys may shape key domains of children's development, including imagination, language, anthropomorphism, relationship formation, epistemic trust, and persuasion. They argue that AI-enabled toys redistribute agency in play, shifting some control over narrative, dialogue, relationship, and information from the child alone to the child-toy interaction. This redistribution may constrain play when the toy functions as an organizing force, but may enable play when the toy serves as a responsive resource for child-led imagination, inquiry, and social engagement. The authors conclude by outlining implications for developmental research, product evaluation, and child-centered AI design.

Key Findings

1. AI-enabled toys are increasingly entering children's everyday play, yet their developmental implications remain poorly understood.
2. The paper reviews potential effects across imagination, language, anthropomorphism, relationship formation, epistemic trust, and persuasion.
3. The central mechanism proposed is a redistribution of agency in play: some control over narrative, dialogue, relationship, and information moves from the child to the child-toy interaction.
4. Whether this redistribution is harmful or beneficial depends on the toy's role — constraining when it organizes play, enabling when it responds to child-led imagination, inquiry, and social engagement.
5. The authors call for more developmental research, product evaluation, and child-centered AI design grounded in these mechanisms.

Discussion

The paper brings a childhood-development perspective to the AI-in-education literature, extending the conversation beyond formal schooling into the everyday play environments where children increasingly encounter AI. Its framing of agency redistribution — distinguishing a toy that organizes play from one that responds to it — offers a design-relevant lens for pedagogical agents and educational robotics more broadly. By connecting AI toy design to child-centered development (imagination, language, attachment, trust), it foregrounds the Learner Agency and Well-Being considerations that formal AI Literacy curricula rarely reach. For the knowledge base, it broadens the AI in Education domain toward early-childhood and informal learning, and complements the literature on conversational agents and embodied AI in education.

What this means for practice

• Designers. Build toys that respond to child-led imagination rather than organize play: agency redistribution constrains play when the toy drives narrative and dialogue and enables it when the toy serves as a resource for the child's inquiry.
• Designers. Preserve space for child-generated dialogue and narrative instead of substituting system-generated content, the mechanism through which earlier talking toys narrowed the range of play themes children produced.
• Designers. Design the interaction to support calibration of epistemic trust in toy output, because children may treat fluent system-generated content as reliable information and because persuasion is a live risk in open-ended conversation.
• Researchers. Study young children directly rather than extrapolating from the self-reported adolescent data that currently carry the relationship-formation and self-disclosure claims.

Limitations

• This is a literature review with no primary data, so its developmental claims are mechanistic inferences from prior work rather than tested outcomes.
• The established evidence base concerns earlier scripted toys — button-press toys, talking dolls, and branching choose-your-own-adventure toys — whose responsiveness differs substantially from open-ended generative conversation, yet those findings carry the arguments about imagination and language.
• For relationship formation the authors state there is no direct evidence on young children: the supporting findings are self-reported studies with adolescents who disclose more to chatbots than to people.
• Systematic empirical work on the qualitative shifts in attachment is limited to illustrative accounts such as Turkle (2007), leaving the risk claims about blurred boundaries between pretend and real social engagement unmeasured.

Citation

Xu, Y., Girouard, L., & Shi, Z. (2026). Artificial Intelligence in Toys: Implications for Child Development and Play. EdArXiv preprint.