Explainable AI for student prompting in higher education

The use of educational technology (EdTech) increased exponentially the past decades. Higher Education institutions (e.g. VIVES) follow this trend. Recent developments in Artificial Intelligence (AI) generate many applications for education.Teaching and learning using EdTech (e.g. using online learning environments) is challenging for teachers and students but has the potential to produce valuable data streams which  can support teaching and learning. One of the applications of using these data is to predict student success.

For example, a VIVES-research team succesfully implemented an AI-algorithm facilitating the prediction of scores for nursing students on an online introductory anatomy course. This model predicted student results ‘accurately’, ‘early’ (i.e. halfway a semester), and ‘explainable’. Explainable prediction of student succes allowed teachers and students to act upon the AI-generetad information to increase the chances of students passing a course.

In this project we aim to: a) capture and combine multiple data streams on student behavior (including on- and offline activity) for a wide range of courses in VIVES. Doing so we build a rich datapool to train AI-models b) Identify meaningful features that serve as ‘explainable predictors’ for student success c) use AI-techniques (e.g. machine learning) to predict student success in a timely, accurate and explainable manner.  d) Use AI output and Large Language Models (LLM) to prompt students to improve their learning behavior and in turn increase succes rates The focus of this project is on generalising ‘explainable AI’.

We aim to implement AI-support across multiple courses in different domains. This means working with different teachers and large student  populations. By generating predictions for student’s performance along with explanations we aim to avoid a black-box phenomenon and increase the trust and confidence students and teachers have in AI-supported education. This project aims for high quality, AI-supported education in suitable courses of VIVES University of Applied sciences. Supporting education with explainable AI has the potential to increase student succes rates which, in Flanders, is linked to University funding. Also, we contribute to a scientific research base on using AI in education.

We aim for publication of both scientific and practitioner oriented papers as well and to present our findings to a broad audience. Predicitive AI-models can be packaged, published and licensed to become available for external partners in different domains using EdTech (e.g. higher education, companies using EdTech for training, companies creating EdTech etc …). Finally, VIVES-students can use (anonymized) data to learn from and assist in development.