Just ask Mattie

Supporting first-year business maths students with a custom-built AI-tutor

Abstract

This paper presents preliminary findings from the development and deployment of Mattie, an AI-powered chatbot tutor designed to support first-year university students in a business mathematics course. Motivated by concerns about students’ reluctance to seek help, particularly in public forums where they fear appearing ‘stupid’ in front of peers or instructors, Mattie.AI was created to provide personalised and course-specific learning assistance on demand 24/7. Unlike general-purpose chatbots, Mattie.AI was built using a Retrieval-Augmented Generation (RAG) architecture, allowing it to generate responses based on course-specific learning materials. The underlying chatbot design was Socratic, focused on helping students through guided questioning instead of simply providing answers. Evaluation of the first deployment combined chatbot interaction data with student feedback from a usability and satisfaction survey to provide insights into its functionality as well as patterns of student engagement. Results suggest that subject-specific chatbots offer a promising supplement to traditional support structures, but students need to be supported to use them effectively. Future work in this space of AI-enhanced education technologies will refine Mattie’s capabilities and explore its impact on learning outcomes and help-seeking behaviours across different student cohorts.