Duolingo dissertation grant supports research into AI-generated mathematical stories

Mathematical problem-solving can be challenging for many children, particularly when abstract ideas and unfamiliar language make problems difficult to understand. Our research explores whether storytelling can make this process more meaningful and engaging by placing mathematical challenges within relatable narratives.

MathsTale: a personalised approach

We are investigating how generative AI can support this approach by creating personalised mathematical stories that respond to children’s different needs and interests. As part of this work, we have developed MathsTale, a generative AI-powered web application designed to help primary-school children learn mathematical problem-solving through interactive stories. Our earlier studies with children and educators have produced encouraging feedback and have helped shape the system’s design.

The next stage of our research will involve refining MathsTale through further user studies, followed by a longer-term evaluation of how the approach affects children’s understanding, engagement and problem-solving experiences. We will also work closely with educators to explore how AI-generated stories could support teaching and learning in practice, and to understand how this technology could be introduced into different teaching practices in a responsible and educationally valuable way. The findings will also contribute to future technical publications and presentations at international conferences.

The Duolingo grant will support the server and cloud-hosting costs required to run MathsTale and keep the platform accessible to participants during our studies. It will also help cover the practical costs of conducting user studies with children and educators, including participant compensation, equipment hire, and access to suitable venues or classroom spaces. 

Research environment

Beyond the direct support provided by the grant, this work also reflects the supportive and interdisciplinary environment of the EPCC PhD programme and our collaboration with the University of Edinburgh School of Informatics. 

EPCC’s computing resources, including the Edinburgh International Data Facility (EIDF), have played an important role by giving us access to GPU-enabled infrastructure for training and evaluating models, as well as running and deploying the large language models used in MathsTale. 

Together, this technical infrastructure and interdisciplinary expertise have enabled us to bring together research in generative AI, human-computer interaction, and education. Ultimately, we hope the research will provide meaningful evidence and practical guidance for designing AI-supported learning tools that are engaging, pedagogically grounded and useful to both learners and educators.

Recent papers

Zhang, K., Wang, H., Niu, Y., Inglis, A. L., Palansuriya, C., & Constantin, A. (2026, June). Interactive MathsTale: A Generative AI-Powered Multimodal Storytelling System with Hands-on Activities for Children’s Mathematical Problem Solving. In Proceedings of the 25th Annual ACM Interaction Design and Children Conference (pp. 1080-1086).

Zhang, K., Niu, Y., Alqassim, M. Y., Inglis, A. L., Palansuriya, C., & Constantin, A. (2026, April). MathTales: Designing and Studying an AI Agent-Based Story Generation System for Teaching Mathematical Problem Solving to Children. In Proceedings of the Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (pp. 1-13).

Authors

Kejia Zhang, PhD student, EPCC; Dr Charaka Palansuriya, EPCC; Dr Aurora Constantin, School of Informatics, University of Edinburgh.