Multicultural Interactive Integration Using Social Agents (MI2US)
MI2US uses social robots and language models to help children from multicultural backgrounds feel included and thrive in school settings.
Duration: July 2023 – December 2024
Status: Completed
Educational Level: Primary Level, Lower Secondary Level
Topic: Artificial Intelligence AI, Digital Tools
Keywords: Adaptive Learning System, Digital Learning Games, Educational Robotics, Artificial Intelligence, Learning Platforms, VR – Virtual Reality
Initial Situation
Adapting to school life in Switzerland can be challenging for children from multicultural backgrounds. On top of the always challenging language barriers, other topics also add complexity to these problems that are mostly handled by teachers. Unfamiliar social norms, lack of cultural representation, and lack of self-confidence often lead to feelings of isolation and reduced participation. These challenges can hinder both academic progress and emotional well-being. Social robots, powered by large language models, offer a promising solution. They can engage children in multilingual, culturally sensitive interactions, provide emotional support, and help bridge communication gaps. By creating a safe and inclusive space, these robots foster confidence, encourage social integration, and support learning, making classrooms more welcoming for every child.
Objectives
MI2US is an innovative initiative that harnesses the power of social robots and large language models to support the integration of children from diverse cultural backgrounds into educational environments. By combining empathetic AI-driven interactions with inclusive pedagogical strategies, MI2US creates engaging, supportive spaces where every child feels seen, heard, and valued. The project aims to foster belonging, boost confidence, and promote cross-cultural understanding, laying the foundation for more inclusive classrooms and communities.
Method
First, we interviewed teachers and special needs teachers from the cantons of Vaud and Bern to gather their opinions on using different technologies for multicultural integration in their classrooms. In parallel, we conducted a systematic literature review on such methods in primary schools. We combined these findings to develop a qualitative synthesis with guidelines for best practices in integrating children with migration backgrounds using interactive technologies in classrooms. Participants in our experiment were 5 special needs teachers from five different public schools in the canton of Bern, up to 15 teachers from private schools in the cantons of Vaud and Geneva, and approximately 150 primary school students.
Results
The main finding suggested that the optimal use of technology in this context is to provide hypothetical training scenarios for children, enabling them to express their own culture (and their feelings about it) while being exposed to each other’s cultures. Examples of these scenarios include role-playing and storytelling activities. Based on these insights, methods involving Large Language Models (LLMs) with intuitive graphical user interfaces for teachers were implemented for these activities.
Implemented Translation
We implemented a virtual avatar of the robot that integrates with the software we developed. The software is designed to be accessible to any teacher and adaptable to any topic. By generating integrative storytelling content using a Large Language Model (LLM) trained on educational datasets, the avatar delivers the prompted content while also creating opportunities to reflect on diverse cultural backgrounds and foster their integration. The code can be shared by sending an email to the indicated main contact person. We validated these methods through four experiments involving more than 100 children and six teachers from the cantons of Vaud and Geneva. Children and teachers rated the activities with the robot as more enjoyable and engaging compared to their regular activities. Results can be checked in the publications produced by this project.
