Abstract
Pedagogical agents are typically designed to take on a single role: either as a tutor who guides and instructs the student or as a tutee that learns from the student to reinforce what he/she knows. While both agent-role paradigms have been shown to promote student learning, we hypothesize that there will be a heightened benefit with respect to students’ learning and emotional engagement if the agent engages children in a more peer-like way — adaptively switching between tutor/tutee roles. In this work, we present a novel active role-switching (ARS) policy trained using reinforcement learning, in which the agent is rewarded for adapting its tutor or tutee behavior to the child’s knowledge mastery level. To investigate how the three different child–agent interaction paradigms (tutee, tutor, and peer agents) impact children’s learning and effective engagement, we designed a randomized controlled between-subject experiment. Fifty-nine children aged 5–7 years old from a local public school participated in a collaborative word-learning activity with one of the three agent-role paradigms. Our analysis revealed that children’s vocabulary acquisition benefited from the robot tutor’s instruction and knowledge demonstration, whereas children exhibited a slightly greater effect on their faces when the robot behaves as a tutee of the child. This synergistic effect between tutor and tutee roles suggests why our adaptive peer-like agent brought the most benefit to children’s vocabulary learning and affective engagement, as compared to an agent that interacts only as a tutor or tutee for the child. This work sheds light on how fixed role (tutor/tutee) and adaptive role (peer) agents support children’s cognitive and emotional needs as they play and learn. It also contributes to an important new dimension of designing educational agents — actively adapting roles based on the student’s engagement and learning needs.
Keywords
Intelligent tutoring systems, Interactive learning environments, Evaluation of CAL systems, Cooperative/collaborative learning
Introduction
Early childhood is a critical period of development that sets the foundation for children’s future academic success and aspirations. Unfortunately, only about 30% of eligible 4-year-old children are enrolled in state pre-K programs every year (National Institute of Early Education Research, 2018). Many young children do not have access to quality preschool programs or equivalent homeschooling, and consequently do not achieve kindergarten readiness prior to entering the formal education system (Nores & Barnett, 2014; U.S. Department of Education, 2015). Statistics show many children who start off below readiness level have a hard time catching up (Garcia & Weiss, 2017). Access to extracurricular support (e.g., after-school or summer programs) could help reduce this gap, but resources are limited and can be very costly (Grossman, Lind, Hayes, McMaken, & Gersick, 2009).
In at-risk communities, it is very challenging for kindergartens to offer a curriculum that is cognitively and academically leveled to every student in their classrooms. Children enter school with a wide range of cognitive and pre-literacy starting points, as each child has a unique distribution of the various cognitive, visual, social, and linguistic skills needed to be a successful reader (Dehaene, 2009; Wolf & Gottwald, 2016). Hence, there is a real need and compelling opportunity to develop adaptive educational technologies to supplement the learning experiences that diverse learners receive at school and augment early childhood education — especially given their promise to deliver personalized education and be cost-effective at scale.
A wide variety of technological interventions, such as intelligent tutoring agents, game apps and computer simulations, have been designed to support students across a range of ages and in a variety of academic domains (Belpaeme, Kennedy, Ramachandran, Scassellati, & Tanaka, 2018; Breazeal, Morris, Gottwald, Galyean, & Wolf, 2016; D’mello & Graesser, 2013; Lindgren, Tscholl, Wang, & Johnson, 2016; Vaala, Ly, & Levine, 2015). Inspired by Blooms 2-sigma effect (Bloom, 1984), intelligent tutoring systems (ITS) are a well-established field where the computer instructs, provides feedback, and guides student learning in 1:1 interaction (Graesser, Hu, & Sottilare, 2018). ITS has investigated diverse cognitive elements to support student’s needs encompassing assessing mastery (Baker et al., 2010), modeling student’s cognitive states (Corbett, Kauffman, Maclaren, Wagner, & Jones, 2010), adapting content to individual student needs (Manickam, Lan, & Baraniuk, 2017), as well as looking at student effect and motivation (D’mello & Graesser, 2013). In this learning paradigm, the computer or intelligent agent interacts with students as their tutor or teacher. Alternatively, teachable agents have also been designed to emulate a younger, less capable playmate. This approach is grounded in a widely studied and practiced concept in education called ‘‘learning-by-teaching’’ (Roscoe & Chi, 2007) that enables children to improve and consolidate their learning by teaching another (Biswas, Leelawong, Schwartz, Vye, & at Vanderbilt, 2005; Brophy, Biswas, Katzlberger, Bransford, & Schwartz, 1999; Chin, Dohmen, & Schwartz, 2013; Park & Howard, 2014; Tanaka & Matsuzoe, 2012).
To date, pedagogical agents are mostly designed to serve a single role: either as a tutor or as tutee. However, we argue that a flexible interaction paradigm, in which an educational agent can adaptively switch between roles at appropriate times, holds great potential to leverage the benefits of both as human peers often provide to one another. Given the promise of AI agents for early childhood education and the different interaction paradigms that are possible, it is important to understand how different designs influence young children’s learning and emotional experience. Resulting insights will help inform the design of effective and emotionally engaging educational interventions that support the diverse cognitive, social, emotional, and physical learning needs in early childhood. The main research questions in this paper are (1) how do young children learn or engage when interacting with different types of pedagogical agents, and (2) how can we leverage the benefits of each pedagogical agent paradigm to provide a synergistic impact on young children’s learning and engagement. More specifically, we are particularly interested in comparing different child–agent interaction paradigms where the agent takes on a different role (i.e., tutor, tutee, or reciprocal peer) and how each impacts children’s learning and effective experience.
Our work makes the following contributions. First, we introduce a new agent-role paradigm where the educational agent acts as a reciprocal peer. Tutor agents, designed to explicitly teach students via instruction, demonstration, and feedback, is a well-established paradigm. More recently, pedagogical agents that act as a tutee to engage children in a learning-by-teaching paradigm have been proposed. To our knowledge, no prior work studied how a pedagogical agent switched its role between tutor and tutee when interacting with children. In this work, we designed a novel adaptive role switching (ARS) model whereby the robot can flexibly change its role in a reciprocal interaction to teach and learn from each child. We developed the ARS model using reinforcement learning to maximize children’s exposure to both tutor and tutee roles at appropriate times. We first pre-trained the model using a pilot study dataset where the robot randomly switched between roles as it played the vocabulary game with children. We used this baseline model to seed the ARS policy training for the model’s faster convergence and adaptation to each child in our main study. The robot, in effect, learned when to switch its role to stay in sync with each child’s real-time learning performance.
Second, our work is the first experiment to our knowledge that directly compares the impact of a tutor, tutee, or reciprocal peer robot on young children’s learning and effective engagement. We designed a between-subjects experiment with 59 children aged 5–7 years old divided into three counterbalanced groups, and assigned to one of our three experimental conditions. An expressive and appealing social robot, Tega, was used as the learning companion in our work (Fig. 1(a)). In the tutee condition, the robot behaved as a curious learner who lacked vocabulary knowledge and needed the child’s help. In the tutor’s condition, the robot behaved as an expert that never made mistakes and always gave the child feedback and guidance. In the peer condition, the robot used the ARS policy to determine which role to exhibit at each turn of gameplay. Participants played a vocabulary learning game with a Tega robot for two 30-min sessions, learning new target vocabulary in each. Each group was compared and evaluated with respect to children’s word learning performance and facial expressions as an indicator of effective engagement. We found that children’s vocabulary learning and affective behavior were the most enhanced with the reciprocal, adaptive peer robot.
Our analysis found that children’s vocabulary acquisition benefited from the robot’s instruction and knowledge demonstration, whereas children’s facial effect was slightly more expressive when the robot behaved as a tutee of the child. This synergistic effect between tutor and tutee roles suggests why our adaptive peer-like agent brought the most benefit. Hence, our third contribution is a novel user experience paradigm for peer-like educational agents that can successfully engage children in reciprocal and adaptive tutor–tutee roles. In light of these findings, we provide design guidance for effective and engaging peer-like learning companions that promote the growth and welfare of young children.
Conclusion
This work is the first to implement and evaluate a reinforcement learning-based reciprocal child–agent peer-learning paradigm for children and is the first to compare the impact of different roles of pedagogical agents tutor, tutee, peer on children’s learning and affect together. We developed a novel bidirectional child–agent peer-learning paradigm, inspired by children’s peer-to-peer interaction and built using a reinforcement learning model. The robot was rewarded for synchronizing its behaviors to a child’s knowledge level, and for its scaffolding actions when the child is struggling. This enabled the child and robot to be exposed to both tutor and tutee roles dynamically, creating a socially rich interactive experience that builds a sense of camaraderie. In fact, in the videos we annotated, we do see children positively responding to the robot’s verbal and nonverbal encouragement cues and offering the same to the robot, evidencing their emotional and relational engagement with the peer robot. We explored the impact of different educational agent’s roles (tutor, tutee and peer) on children’s learning and emotional engagement through collaborative play in the context of an educational game. We found that children who interacted with a reciprocal, adaptive peer agent showed the greatest vocabulary learning, varied face-based effect, and positive valence among the three types of pedagogical agents. In sum, our technical contribution and real-world evaluation study in a public school add to a growing body of work exploring how different student–agent interaction paradigms impact young children’s behavior, emotional engagement, and learning. Given the importance of effective interventions during early childhood and the importance of creating an emotionally engaging experience for young children, we hope this work contributes to the realization of intelligent and emotionally engaging pedagogical agents for this important, yet relatively under-served learner population with AI-enabled solutions.
About KSRA
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:
Teaching and learning with children: Impact of reciprocal peer learning with a social robot on children’s learning and emotive engagementBibliography
author
Year
2020
Title
Teaching and learning with children: Impact of reciprocal peer learning with a social robot on children’s learning and emotive engagement
Publish in
Computers & Education, Volume 150, June 2020, 103836
Doi
https://doi.org/10.1016/j.compedu.2020.103836
PDF reference and original file: Click here
Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.
Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.
