Learning with Conversational Agents
Skip to main content
eScholarship
Open Access Publications from the University of California

UC Irvine

UC Irvine Electronic Theses and Dissertations bannerUC Irvine

Learning with Conversational Agents

  • Author(s): Xu, Ying
  • Advisor(s): Warschauer, Mark
  • et al.
No data is associated with this publication.
Abstract

This dissertation consists of four studies that reveal how children learn from, respond to, interact with, and perceive conversational agent as their learning companion during shared storybook reading. The first study focuses on how children learn from the agent. I found that contingent, structured dialogue with the conversational agent led to children’s enhanced story comprehension. Such benefit was largely driven by children’s heightened level of vocalizations related to the story narratives and reduced off-topic vocalizations. The second study primarily focuses on how children respond to the conversational agent. I found that conversational agents promoted children’s response intelligibility, while adults elicited longer, more lexically diverse, and more relevant responses. The differences in language productivity were amplified among the questions requiring high cognitive demand. The third study focuses on how children interact with the agent verbally and non-verbally. I found that children generally participated in the conversation with the agent smoothly: they generated on-topic responses and answered within the proper time frame. The result also confirmed the advantage of using a combination of open-ended questions as initial prompts to encourage children’s free expression and multiple-choice questions as follow-up prompts to help ease the potential cognitive obstacles. Such scaffolding mechanisms appeared to benefit younger children more so than older ones. The fourth study focuses how children perceive the agent. I found that children in general held positive perceptions in terms of conversational agents’ cognitive and psychological capabilities. Children’s such perceptions establish the feasibility of developing agents to socially engage children in learning activities. Overall, the four studies provide converging evidence on the promise of leveraging AI-powered conversational technologies to support young children’s language development. The findings are intended to be generalized to designing socially interactive environments for different learning domains (e.g., science) and learning scenarios (e.g., television watching), with a goal of promoting children’s long-term development.

Main Content

This item is under embargo until June 29, 2022.