Digital Twin Universities: Creating Real-Time Virtual Replicas of Students for Predictive and Personalised Higher Education

Abstract- The intersection of Artificial Intelligence (AI), learning analytics, and digital twin is revolutionising higher education into an intelligent, data-driven, and ultimately personalised educational ecosystem. In contrast with today’s Learning Management Systems (LMSs), which offer reactive, retrospective perspectives and generalised student characteristics, a digital twin in education provides real-time, personalised, and proactive learning guidance for each student based on the learner’s real-world educational activities. To address this shortcoming, this article reports on a DTU framework for modelling and maintaining an individualised, evolving twin for each student in an online version of a real university. The individual twin incorporates multimodal educational information describing the learner’s engagement and behaviour, state of mind, learning performance, and employment-related skills. The proposed DTU layered design will intelligently monitor students' status in real time, proactively predict students' academic performance and current level of engagement, identify students at risk of dropping out, and dynamically generate a personalised learning pathway for each student. To validate the feasibility of the proposed DTU framework, we conducted a rigorous controlled quantitative experiment with a sample of 386 students over an entire academic semester in a real-world university setting. Our empirical evaluation results suggest that DTU significantly outperforms several popular LMSs in predictive accuracy (achieving the current best accuracy of 94.6% for student performance prediction) and in improving students' learning gains, engagement, and retention. Early results from the institutional-level experiment also indicate that the DTU has great potential to be a powerful next-generation tool for facilitating personalised and proactive interventions in an AIenhanced university. It has proved to be a solid theoretical concept and an empirically verified methodology for scalable, real-time, personalised learning in modern universities.