Edited by: Mila Vulchanova, Norwegian University of Science and Technology, Norway
Reviewed by: Ramesh Kumar Mishra, University of Hyderabad, India; Vera Kempe, Abertay University, UK
*Correspondence: Paul Vogt
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
In this digital age social robots will increasingly be used for educational purposes, such as second language tutoring. In this perspective article, we propose a number of design features to develop a child-friendly social robot that can effectively support children in second language learning, and we discuss some technical challenges for developing these. The features we propose include choices to develop the robot such that it can act as a peer to motivate the child during second language learning and build trust at the same time, while still being more knowledgeable than the child and scaffolding that knowledge in adult-like manner. We also believe that the first impressions children have about robots are crucial for them to build trust and common ground, which would support child-robot interactions in the long term. We therefore propose a strategy to introduce the robot in a safe way to toddlers. Other features relate to the ability to adapt to individual children’s language proficiency, respond contingently, both temporally and semantically, establish joint attention, use meaningful gestures, provide effective feedback and monitor children’s learning progress. Technical challenges we observe include automatic speech recognition (ASR) for children, reliable object recognition to facilitate semantic contingency and establishing joint attention, and developing human-like gestures with a robot that does not have the same morphology humans have. We briefly discuss an experiment in which we investigate how children respond to different forms of feedback the robot can give.
Given the globalization of our society, it is becoming increasingly important for people to speak multiple languages. For instance, the ability to speak foreign languages fosters people’s mobility and increases their chances for employment. Moreover, immigrants to a country need to learn the official host language. Since young children are most flexible at learning languages, starting second language (L2) learning in preschool would provide them a good opportunity to acquire the second language more fluently at a later age (Hoff,
One trend in the digital age of the 21st century is that technologies are being developed for educational purposes, including technologies to support L2 tutoring. There exist many forms of digital technologies for PCs, laptops or tablet computers that support second language learning, although there is little evidence about their efficacy (Golonka et al.,
Recently, an increasing interest has emerged to develop social robots to support children with learning a second language (Kanda et al.,
While there is a fair body of research on robot tutors, a comprehensive description of the design features for a second language robot tutor based on what is known about children’s language acquisition is lacking. What are the design features of child-robot interactions that would support second language learning? And, to what extent can these interactions be implemented in today’s social robot technologies? In this perspective article, we try to answer these questions based on theoretical accounts from the literature on children’s language acquisition in combination with our own experiences in designing a tutor robot.
In our project, we aim to design a digital learning environment in which preschool children interact one-on-one with a social robot that supports either their learning of English as a foreign language, or the school language for those children who have a different native language (Belpaeme et al.,
One of the first questions that comes up when designing a robot tutor is whether the robot should take the role of a teacher or a peer. Research on children’s language acquisition has demonstrated that children learn more effectively from an adult who can use well-defined pedagogical methods for teaching children using clear directions, explanations and positive feedback methods (Matthews et al.,
To implement effective tutoring, the robot needs to interact with children in multiple sessions, so they have to be motivated to engage in long-term interactions with the robot. Establishing common ground between child and robot can contribute to this (Kanda et al.,
Despite the wealth of studies regarding the introduction of entertainment robots as toys to children (e.g., Lund,
Pilot studies revealed that some children got anxious when the robot was introduced and then suddenly started to move. To familiarize children prior to their first encounter with the robot, it is therefore advisable to prepare them well. For our study, we sent coloring pages of the robot to the preschools during recruitment and asked the pedagogical assistants to talk a little bit about the robots to the children. About 1 week before the experimental trials, the experimenters introduced the robot in class during their daily “circle time”, as this provided a safe and familiar environment with the whole group in which the pedagogical assistants usually introduce new topics or new activities. One experimenter first introduced the robot by telling a story about Robin, the name of our robot, using a makeshift picture book. In this story we explained the similarities and dissimilarities between the robot and children to construct the type of common ground considered to have a positive effect on the learning outcome (Kanda et al.,
After this story, another experimenter entered the room with the robot while it was actively looking at faces to provide an animate feeling. The robot introduced itself with a small story about itself and by performing a dance in which the children were encouraged to participate. The end of the circle time consisted of getting a blanket for the robot so it could “sleep”. This introduction was repeated later on the days we conducted the experiment in one-on-one sessions. While by then most children were comfortable interacting with the robot, some were still timid and anxious. To encourage these children to feel comfortable, one of the experiment leaders would sit next to the child during the warm-up phase of the experiment and motivate the child to respond to the robot when necessary until the child was sufficiently comfortable to interact with the robot by herself/himself. We found that the younger 3-year olds required more support from the experimenters than the older 3-year olds (Baxter et al.,
Research has shown that it is crucial for children’s language development that their communication bids are responded to in a temporally contingent manner (Bornstein et al.,
Robots should not only respond to children in a timely fashion, but also in a semantically contingent fashion (i.e., consistent with the child’s focus of attention), as this too has a positive effect on children’s language acquisition (Bornstein et al.,
To understand children’s communication bids, as well as to test their pronunciation of the L2, it is important that the robot be equipped with well-functioning automatic speech recognition (ASR). However, the performance of state-of-the-art ASR for children is still suboptimal, especially for preschool-aged children (Fringi et al.,
In our project, we explore various strategies to achieve this, both based on monitoring non-verbal behaviors of the children and focusing on comprehending rather than producing L2. The first strategy relies on providing children tasks they have to perform in the learning environment, such as placing “a toy cow behind a tree” when teaching spatial language. This, however, requires the visual object recognition on the robot to work well, which is only the case when the scene contains a limited set of distinctively recognizable objects, such as distinctly colored objects (Nguyen et al.,
Joint attention, where interlocutors attend on the same referent, is a form of social interaction that has been shown to support children’s language learning (Tomasello and Farrar,
Feedback, too, is an interactional feature known to help language learning (Matthews et al.,
To investigate how children experience feedback from a peer robot, we carried out an experiment among 85 3-year-old Dutch-speaking children at preschools in Netherlands (de Haas et al.,
Finally, from a pedagogical point of view it is desirable that the interactions between child and robot be sufficiently challenging and varied so that the child has a target to learn from, but at the same time interactions should not be too difficult, because that may frustrate the child causing it to lose interest in the robot (Charisi et al.,
This perspective article presented some design features that we consider crucial for developing a social robot as an effective second language tutor. We believe the robot is most effective when it is framed as a peer, i.e., as a fellow language learner and playmate, but that is designed to use adult-like interaction strategies to optimize learning efficacy. In order to establish common ground and trust to facilitate long-term interactions, we consider it essential that the robot be introduced with appropriate care on the first encounter. As an example, we outlined our strategy for introducing a robot to preschool children. Interactions between child and robot should be contingent and multimodal, and provide appropriate forms of feedback. We argued that the robot should remain within Vygotsky (
We also discussed some technical challenges that need to be solved in order to implement contingent interactions; the most important of which we believe is ASR, which presently does not work well for children’s speech. While various technical challenges still remain, we expect that social robots will provide effective digital technologies to support second language development in the years to come.
The present list of design features covers many aspects that need to be considered when developing a tutor robot, but it is not yet comprehensive. One aspect that has not been covered, for instance, concerns the design of robots for children from different cultures, which could require different design choices (Shahid et al.,
The Research Ethics Committee of Tilburg School of Humanities approved this study, and the parents of all participating children gave written informed consent in accordance with the Declaration of Helsinki.
PV, MH and EK designed the conceptual aspects of the article; PV, MH, CJ and PB carried out the literature review; PV, EK and MH designed the feedback study; MH, CJ and PB designed the introduction study; MH, CJ and PB carried out the studies; PV and MH wrote the article; CJ, PB and EK revised the article critically.
This work has been supported by the EU H2020 L2TOR project (grant 688014). CJ and PB thank the research trainee program of the Tilburg School of Humanities for their support.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors wish to thank all members of the L2TOR project for their support and advice regarding this research. We also thank Kinderopvanggroep Tilburg and all participating daycare centers and preschools for their assistance in this research. Finally, a big thank you to all the children and their parents for participating in our research.