Intelligent Tutoring System (ITS) for Pre-University Mathematics

Duration: May 2022 – December 2022
Status: Completed
Educational Level: Upper Secondary Level – Grammar School Education
Topic: Digital Tools
Keywords: Adaptive Learning System

Initial Situation

Over the past decade, computer-based learning—including e-learning, mobile learning, educational games, and standalone applications—has grown significantly due to technological advances and the shift toward distance learning during the Covid-19 pandemic. However, these systems still lack the flexibility, feedback, and personalisation of real-life classrooms. Adaptive learning systems, which tailor content to each learner’s needs, aim to address this gap. Central to this approach are Intelligent Tutoring Systems (ITS) that rely on student models to trace knowledge, predict behaviour, and personalise learning paths. Despite promising developments, practical implementations remain rare, largely due to the difficulty of applying general-purpose ITSs across various subjects and age groups.

Objectives

This project aimed to respond to that challenge by developing an ITS specifically for teaching Swiss high school mathematics (ages 16–18), a context where adaptive systems have not yet been widely adopted. It also sought to collect data for AI-powered knowledge tracing, with a particular emphasis on identifying blocking states and learning moments.

Method

The team collaborated with Taskbase – a Swiss start-up with prior experience in adaptive learning platforms – to accelerate the development process. A web-based interface was designed to enable the uploading of exercises, the provision of explanations, and data collection, and an initial prototype was developed. The web application was planned to be deployed in upper secondary schools (Gymnasien) in the Canton of Bern, in collaboration with PHBern, to collect data from students. The knowledge tracing approach was intended to combine probabilistic and deep learning methods. To identify blocking states, user interactions on the interface (e.g., help clicks, errors, waiting times) were to be analysed. The collected data were then to be examined to determine whether and when learners had mastered specific skills or concepts.

Results

As part of the project, an initial prototype of a web-based ITS for uploading exercises was developed. The ITS prototype enabled the collection of valuable data for knowledge tracing. The project focused on identifying blocking states and learning moments. These aspects had previously been under-researched due to detection challenges. Early data allowed the team to explore models that can predict such states and refine adaptive features in the system.

Implemented Translation

Collaboration with Taskbase enabled efficient development based on proven adaptive learning technologies. However, due to the project’s limited budget, it was not possible to include a sufficient number of students or to implement the full range of planned functionalities. The intended deployment in Bernese high schools and the associated data collection for the further development of knowledge-tracing models could not be realized. Nevertheless, the project provided important conceptual foundations: the approaches developed for detecting blocking states (moments when learners get stuck) and learning moments (points of knowledge acquisition), as well as the design of an ITS (Intelligent Tutoring System) for advanced mathematics topics within the Swiss high school curriculum, form a valuable basis for future projects that could be continued with sufficient resources. The project made a conceptual contribution to the development of adaptive learning systems for advanced mathematics. By focusing on the Swiss high school curriculum and the 16–18 age group, it addressed a gap in existing research, which has largely relied on North American datasets and younger students. The methodological approaches developed for identifying blocking states and learning moments provide theoretical foundations for more precise analyses of learning processes in adaptive systems. Although budget constraints prevented practical implementation and data collection, the project documents both the potential and the challenges of developing personalized digital learning aids for demanding mathematical topics. It provides a conceptual foundation for further research and development in adaptive learning systems at the upper secondary level.

Participating institutions