A number of studies have found that children engage in natural conversation with artificial intelligence (e.g., robots, voice assistants) which indicates the feasibility of voice agents as social partners for children. Moreover, dialogic reading, which includes asking open-end questions to stimulate children’s thinking and providing feedback, has been identified as amplifying the benefits of storybook reading for children. Xu and colleagues compared the effects of dialogic reading with a human and dialogic reading with a non-human agent on promoting children’s language skills.
Researchers recruited 117 children (mean age = 58.1 months; 31% Asian) from five childcare centers serving middle-class communities in US and data were collected from Feb to Aug 2019. A two-by-two factorial design was adopted with human vs agent with voice only interface (Google Home Mini device), and dialogic reading (i.e., conducting the narrative reading and engaging children in dialogue by asking questions and providing feedback) vs non-dialogic reading (merely narrating the same story). Children were randomly assigned into four conditions:
- Agent Dialogic Reading (Agent DR)
- Agent Non-Dialogic Reading (Agent Non-DR)
- Human Dialogic Reading (Human DR)
- Human Non-Dialogic Reading (Human Non-DR)
Both human and agent followed the same dialogue script for the two dialogic reading groups. Children’s baseline vocabulary skills were accessed using standard test. The research team developed a test battery which was used to measure children’s story comprehension. Children’s engagement while listening to the story was coded from video-taped reading sessions. After controlling for pretest vocabulary skills, analysis results indicated that:
- For overall story comprehension, the dialogic reading condition, compared to the non-dialogic reading, led to significant effect (ES = +0.51), while agent and human had no significant difference (ES = -0.14).
- Human or AI agent as reading partner had no significant moderating effect on story comprehension.
- In terms of overall reading engagement, dialogic reading had significant effect (ES = +0.41) while no significant effect was found between reading partners (ES = 0.00).
- In terms of children speaking during a reading episode, dialogic reading increased narrative-relevant vocalization (ES=+1.11) compared to non-DR, while an agent reduced content vocalization irrelevant to the story (ES=-0.63) compared to a human reading partner.
- A possible mechanism was examined: agent dialogic reading promoted narrative-related vocalization (compared to “non-DR” groups) and decreased irrelevant vocalizations (compared to “human” partner), through which, comprehension scores were increased.
The experiment results revealed that a conversational agent, even just voice only (disembodied), replicated the benefit of dialogue with an adult as reading partner. Though researchers did not suggest robots replace parents during children’s story time, conversational agents are a cost-effective tool for enriching preschool-age children’s literacy development.
Source (Open Access): Xu, Y., Aubele, J., Vigil, V., Bustamante, A. S., Kim, Y., & Warschauer, M. (2021). Dialogue with a conversational agent promotes children’s story comprehension via enhancing engagement. Child Development, cdev.13708. https://doi.org/10.1111/cdev.13708