Alice Chambers |
The University of New South Wales (UNSW) in Sydney, Australia, has developed an AI-powered model with Microsoft solutions to detect when students are at risk of academic failure and connect them with the right support and services.
The model is part of a project called ‘Data Insights for Student Learning and Support’ and is powered by Microsoft Azure, Azure Machine Learning Studio, Azure OpenAI Service and Power Apps. To run the model, UNSW brought together data from different departments and systems within one data lake with help from Microsoft and its partners Accenture and Altis.
“The support of Microsoft and their partner Accenture really helped us kickstart everything through the co-development of a prototype in the Power Apps Innovation Centre Program,” said Simon McIntyre, director of education innovation at UNSW, in a Microsoft blog post. “Our chief data and insights officer and Altis then helped wire custom configurations [of our Microsoft technology stack] together, which we wouldn’t have been able to do as quickly on our own.”
The model was first tested on a small scale in 2023 with 33 academics and 25 courses. UNSW found that the model accurately identified 79 per cent of at-risk students in the first few weeks of a course.
Testing then expanded in early 2024 to 80 courses, which included approximately 17,000 students and 83 academics. The model identified 284 students at risk of failing and helped academics to provide them with support. The pilot also enabled 75 per cent of academics to identify potential failure risks much earlier than previously possible and 49 per cent of students who received proactive help showed increased class engagement.
“Being on this project has really enabled me as a lecturer in charge to understand some things about my students in a quantifiable way,” said Lynn Gribble, associate professor from the school of management and governance at UNSW. “We know that students who perhaps leave something to the last minute, go missing from Moodle [UNSW’s learning management system], or aren’t engaging with the course materials will not do as well as students who do.”
The model is set to roll out to all first-year students and teaching staff at the start of 2025 and then reach all 80,000 students and 7,000 staff by the following year.