Conversations With Media Characters Promote Children's Science Learning From Television Watching

• In Event: Using Conversational Artificial Intelligence to Promote Active and Engaging Learning

Abstract

*Objective*

Video programs are important educational resources, and their potential can be amplified if children are allowed to interact with media characters during their video watching. Partnering with PBS KIDS, we developed interactive science videos based on a popular show, Elinor Wonder Why. In these videos, the main character, Elinor, powered by a conversational agent, engaged children in conversation by asking children questions and providing responsive feedback. To examine the impacts these conversational videos have on children’s learning, we conducted two experiments involving a total of 330 children, illuminating the feasibility and effectiveness of using conversational technologies to maximize the educational benefits of television watching.

*Theoretical Framework*

While studies have recognized the benefits of educational video programs, video programming is traditionally a one-way transmission medium that does not allow for the kind of contingent interaction known to be vital for children's engagement and learning (Roseberry et al., 2014; Troseth et al., 2006). Such contingent interaction may be particularly important for science-oriented videos. Studies have suggested that through dialogic interaction, children are exposed to scientific language, and verbal engagement and exchanges that lead to active, mindful, and reflective learning experiences (Alper et al., 2018).

*Method*

In the first RCT, we randomly assigned 77 children into either 1) watching one conversational episode or 2) watching the broadcast version of the episode without any interaction (same as it is aired on PBS KIDS). Immediately after video watching, children’s episode-specific science learning was assessed using a 12-item questionnaire.

In the second RCT, we extended the first study by adding a third condition: 3) a pseudo-interactive format where the media character asked children the same questions as the conversational video but did not register children’s responses and only gave generic feedback. A total of 240 children were assigned into one of the three conditions for a total of two episodes, after each of which they received immediate posttests.

*Results*

Across both studies, the agent was able to satisfactorily decipher and interpret children's responses.

In the first RCT, children who watched the conversational video scored 0.33 standard deviations (SD) higher than those who watched the non-interactive, broadcast video (p < 0.05). Post-hoc analysis suggested that those who primarily spoke a language other than English at home benefited more (SD = 0.58) than their counterparts

In the second RCT, children who watched the conversational video scored 0.46 SD higher than those who watched the broadcast video (p < 0.001), and 0.25 SD higher than those who watched the semi-interactive version (p < 0.05).

*Scholarly Significance*

As the time that children spend watching videos increases and the mode of watching shifts to Internet-connected devices, it is imperative to investigate how new forms of video watching may better support learning. Research and design communities should take advantage of evolving conversational technologies to maximize the educational benefit of children's screen time.

Authors

• Ying Xu, University of Michigan
  Presenting Author

• Julian Levine, University of California - Irvine
  Non-Presenting Author

• Valery Vigil, University of California - Irvine
  Presenting Author

• Daniel Ritchie, University of California - Irvine
  Non-Presenting Author

• Trisha Thomas
  Non-Presenting Author

• Carlos Barrera
  Non-Presenting Author

• Michelle Meza
  Non-Presenting Author

• Shan Zhang, Harvard University
  Non-Presenting Author

• Andres Sebastian Bustamante, University of California - Irvine
  Non-Presenting Author

• Mark Warschauer, University of California - Irvine
  Non-Presenting Author