Ying Xu | AI and the Developing Child: Myths, Evidence, and Open Questions

Why It Matters

AI‑enhanced interactive media can measurably improve early learning outcomes, prompting educators and policymakers to shape safe, child‑focused AI standards.

Key Takeaways

• Children treat AI as playmate, not task tool
• AI‑driven PBS Kids shows boost recall and transfer scores
• Kids increasingly answer AI questions over repeated sessions
• Parasocial bond strength unchanged across interaction conditions
• Home trials show feasibility, mixed parental preferences

Summary

Ying Xu’s Stanford seminar examined how preschool‑age children engage with voice‑based artificial intelligence, arguing that this demographic warrants a child‑centered AI approach distinct from adult usage. She highlighted that young learners interact with AI primarily for companionship, play, and informal learning, rather than goal‑oriented tasks, making the developmental window both vulnerable and ripe for impact.

Xu presented three empirical studies, the most extensive involving interactive PBS Kids episodes. In a controlled experiment, children experienced three conditions: a true AI conversational interface, a pseudo‑interactive script, and a traditional broadcast. Regression‑adjusted results showed the AI‑conversation group achieved higher post‑test recall and transfer scores, confirming that genuine dialogue—not merely questioning—enhances learning. Response rates to AI prompts rose over time, while non‑interactive groups disengaged.

A vivid illustration came from a early video of a child asking Google about princesses, underscoring how children attribute agency to voice assistants. Further, a parasocial relationship scale revealed no significant differences in perceived friendship or realism across conditions, suggesting that AI dialogue does not intensify one‑sided emotional bonds beyond existing media attachments.

The findings imply that safely designed, controllable AI—using dialogue trees rather than open‑ended generation—can be scaled into homes, offering an additional educational media option. However, parental attitudes remain polarized, indicating the need for clear guidelines and evidence‑based messaging as AI becomes embedded in early childhood learning environments.

Original Description

About the Seminar:

This talk will focus on the role and impact of artificial intelligence (AI) on children’s cognitive and social development. It will highlight how children interact with, perceive, and learn from AI systems, including how they develop trust in these “AI companions.” The talk will also discuss emerging evidence and open questions regarding how generative AI tools shape children’s curiosity, creativity, and critical thinking. It will conclude with a discussion of how social science researchers can amplify their collective voice to ensure that AI is developed and implemented in ways that are safe and beneficial for children.

About the Speaker:

Ying Xu is an Assistant Professor at the Harvard Graduate School of Education. Her research examines how artificial intelligence may support or hinder children’s cognitive development, academic achievement, and social-emotional well-being. Her work aims to inform evidence-based practices and policies to ensure that AI serves as a positive force in child and youth development while mitigating potential risks. She earned her Ph.D. in Language, Literacy, and Technology from the University of California, Irvine. Prior to joining Harvard, she was an Assistant Professor of Learning Sciences and Technology at the University of Michigan.