Towards Building a Pedagogical Agent that Supports Children’s Exploration and Home Literacy Education

Abstract

Early childhood education remains a critical topic and challenge in the past decades due to its impact on children’s future. Since the late 90s, early childhood and developmental psychology researchers have started promoting a new child-centered, exploration-and-play-focused early childhood education method. Despite the research findings, most countries still employ early childhood curricula focusing on school readiness and competency in kindergarten. Although, there is a new tendency toward using a holistic approach to early childhood education in institutions to promote child-focused learning and exploration. To support children’s exploration and self-promoted learning outside school, pedagogical agents are an under-studied platform. This thesis investigates the possibility of using a pedagogical AI agent to help children’s exploration at home.

I first describe the design and development of an interactive storybook platform with explorable literacy features, through which children are endowed with the resources to learn by themselves. I then describe the robot’s behavior design for the exploration demonstration with a social robot platform. Later, I discuss two different robot interaction paradigms for the delivery of the demonstration behavior.

I evaluate the system with 35 children and a between-group ABA study design. Participants interacted with the robot for 2 to 4 weeks and completed 8 sessions in total. In one of the study groups, children’s exploration was self-guided with their agency to decide when to interact with the robot peer. The robot re-actively delivered demonstration behaviors in response to the child’s initiation. In the other study group, the robot’s behaviors were driven by its personalization algorithm; thus, it autonomously delivered interactions without the child’s initiation.

The data analyses were conducted on two scales–children’s self-explorative behaviors and vocabulary learning. The result shows that with a proactive robot peer with exploration demonstrations, children adapted to be more explorative than children who interacted with the reactive robot peer, despite that the robot demonstrated exploration in both conditions. Moreover, we find that children’s exploration is associated with their learning in the robot-guided exploration condition, suggesting that children’s self-explorative behavior in the robot-guided group is learning-oriented and related to their learning growth. When comparing children’s adaptation of exploration, we find a ceiling effect common in child-robot interaction–when children exhibited high exploration in the early phases of the intervention, their exploration growth in succeeding intervention sessions is less compared with less explorative children. Finally, we find an association between children’s exploration and the storybook genre, possibly due to their familiarity with the storybook genre and engagement.

In addition, this thesis attempts to understand the educational needs of home literacy programs from parents’ perspectives. After living with the pedagogical social robot for several weeks, the experimenters conducted semi-constructed interviews with the parent. The qualitative analysis and coding of parents’ interview transcripts suggest common themes in robot design and educational functions that the parents want in a long-term pedagogical agent for their home.