Ying Xu, Ph.D.

Assistant Professor
Graduate School of Education
Harvard University

Elinor asks questions

In this project, I am partnering with PBS KIDS to integrate conversational agents into children’s STEM television shows so that children can have contingent interaction with media characters, with the goal of supporting active engagement and learning. The end goal of this project is to distribute the conversational videos as publicly accessible content via PBS KIDS platforms to millions of children across the country.

Children develop their skills and interests in science, technology, engineering, and mathematics (STEM) through informal learning experiences, which, ideally, should be rich with conversation and social interaction (e.g., asking and answering questions). While STEM-oriented television programming has become an important learning resource for many children, its benefits could be amplified by fostering children’s dialogic engagement.

We have been partnering with PBS KIDS to carry out pioneering research that aims to integrate conversational agents into children’s television shows so that children can have contingent interaction with media characters, with the goal of supporting active engagement and learning. I have developed such “conversational videos” for two popular PBS KIDS animated shows, Elinor Wonders Why and Lyla in the Loop, and conducted studies involving children aged four to eight years.

Funding SourcesPermalink

Overdeck Family Foundation (2025-2027): Effects of Using Artificial Intelligence to Support Children’s STEM Learning From Educational Television Programs: Evidence From a Home-Based Intervention NSF: Developing Conversational Videos to Support Children’s STEM Learning and Engagement (2021-2026) NSF: Using Conversational Agents to Foster Preschool Children’s Learning and Engagement from Interactive Science Videos (2019-2022)

HighlightsPermalink

  • Our studies consistently found that the conversational videos elicited children’s active verbal engagement during video watching.
  • These conversational videos improved children’s understanding of STEM concepts, enhanced children’s engagement, and stimulated more science discussion among parents and children compared to videos currently aired on PBS KIDS without conversational characters (paper in Journal of Applied Developmental Psychology ).
  • The conversational videos helped children form a positive relationship with the media characters, which translated into improved learning.

Upcoming StudiesPermalink

  • A between-subject experiment involving 160 children to test the conversational videos of Lyla in the Loop on computational thinking. The study includes both recall of episode-specific content and a transfer component in the post-test, as well as immediate and delayed post-tests.
  • A home-based experiment where conversational videos will be made available to children for one month. This study focuses on the naturalistic setting, extended access, the impact on parent-child interactions, and the feasibility of large-scale distribution.

Findings From Most Recent PapersPermalink

Artificial intelligence enhances children’s science learning from television shows [Editors’ Pick]

The goal of this study was to better understand the benefits of conversational videos compared to pseudo-interactive videos, such as those in Daniel Tiger’s Neighborhood, where the characters simply pose a question, pause, and then provide a generic response. To achieve this, we conducted an RCT with 240 children during the 2021–2022 school year. This RCT Involves a between-subject design with four conditions: conversational videos, non-interactive videos, and pseudo-interactive videos. The key findings were as follows:

  • Learning:

    • Children who watched the conversational videos performed significantly better on a post-test that measured both recall and transfer of learning concepts, with an effect size of d = 0.39 compared to those in the non-interactive condition and an effect size of d = 0.19 compared to those in the pseudo-interactive condition.

      Interactive learning leads to the highest scores across all science learning measures

  • Engagement

    • Children in the pseudo-interactive and conversational video conditions received identical questions. However, the response rate was higher in the conversational video condition (81%) than in the pseudo-interactive condition (69%).

    • Interestingly, as the video progressed, children in the conversational video condition became increasingly inclined to respond to Elinor’s questions, while those in the pseudo-interactive condition were less likely to respond.

      Interactive group shows a higher and increasing response rate compared to pseudo-interactive group

Examining children’s perceptions of AI-enabled interactive media characters

This study focuses on how contingent interactions shape children’s perceptions of media characters, particularly how children form parasocial relationships with the characters and perceive the characters’ capabilities. The findings suggest that interactive characters, which engage viewers through contingent dialogues, are perceived as more human-like. Follow-up analyses also revealed that the opportunities for contingent interactions with the media character were the primary reason that drove such perceptions.

Selected PublicationsPermalink

  • Xu, Y., Thomas, T., Li, Z., Chan, M., Lin, G., & Moore, K. (2024). Examining children’s perceptions of AI-enabled interactive media characters. In International Journal of Child-Computer Interaction [DOI]

  • Xu, Y., He, K., Levine, J., Ritchie, D., Pan, Z., Bustamante, A., & Warschauer, M. (2024). Artificial intelligence enhances children’s science learning from television shows. In Journal of Educational Psychology.[DOI]

  • Xu, Y., Vigil, V., Bustamante, & Warschauer, M. (2022). Contingent interaction with a television character promotes children’s science learning and engagement. In Journal of Applied Developmental Psychology.[DOI]

  • Xu, Y., Vigil, V., Bustamante, & Warschauer, M. (2022). “Elinor’s talking to me!”: Integrating conversational AI into children’s science narrative programming. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. [DOI]

  • Xu, Y., & Warschauer, M. (2020). Using conversational agents to foster young children’s science learning from screen media. In Proceedings of the 19th ACM International Conference on Interaction Design and Children (IDC ’20). [DOI]

  • Xu, Y., & Warschauer, M. (2020). Wonder with Elinor: Designing a socially contingent video viewing experience. In Proceedings of the 19th ACM International Conference on Interaction Design and Children (IDC ’20). Association for Computing Machinery, New York, NY, USA. [DOI] (Research and Design Competition Honorable Mention)

  • Xu, Y., & Warschauer, M. (2020). “Elinor is talking to me on the screen!” Integrating conversational agents into children’s television programming. In CHI Conference on Human Factors in Computing Systems Extended Abstracts. April 25-30, 2020, Honolulu, HI. ACM. [DOI]

Updated: