Researchers infer that continuous monitoring allows for early intervention immediately after signs of risk are first identified. This quick AI-based response system is crucial when preventing students from dropping out.
Researchers infer that continuous monitoring allows for early intervention immediately after signs of risk are first identified. This quick AI-based response system is crucial when preventing students from dropping out.
A team of researchers at the Universitat Oberta de Catalunya has created an artificial intelligence-based system that can identify students at risk of failing, and automatically send personalized messages to intervene and improve their outcomes.
The researchers noted how continuous monitoring allows for early intervention immediately after signs of risk are first identified. This quick response system is crucial when preventing students from dropping out.
The technology was piloted with 581 students in first-semester courses across several of the UOC’s Faculty of Economics and Business bachelor’s degree programs. It reportedly reduced dropout rates and increased student participation throughout the semester.
Building an innovative AI model
The promising results have been achieved through the use of a new predictive model called Profiled Dropout At Risk, which has been integrated into the Learning Intelligent System.
The LIS, also developed by the research team, predicts whether students are at risk of failing and has undergone successful testing in various pilot programs involving UOC students since 2019.
The predictive model determines whether students are likely to complete an academic year. The inputs are based on historical course data gathered in the Datamart of the UOC’s Institutional Project Evaluation Unit, as well as the results of continuous assessment activities conducted during the current academic year.
After each activity, the LIS predicts the minimum grade required for a student to pass the course and assigns a risk level for potential failure. This information is displayed in the student’s personal area using a traffic light system.
The model offers daily predictions of the risks of dropping out. It takes into account students’ characteristics, performance during the academic year, and their daily interactions with the online campus. It assesses whether a student’s daily engagement aligns with the course’s average, allowing for more proactive intervention based on specific courses and activities.
The risk of false positives
But while this predictive system is known to be helpful for students, it has limitations because it relies on specific checkpoints after each activity. This can result in some interventions being initiated too late – that is, when students have already dropped out for the year.
The new model also faces the challenge of avoiding false positives, where students are incorrectly identified as being at risk. To address this issue, the system considers a time window automatically calculated based on the academic year and types of activities undertaken by the student. This method accounts for the varying levels of student activity within the virtual learning environment.
Empowering students and the teaching staff
Intervention by the system aims to boost students’ motivation through various means. This includes providing recommendations on effective time management, setting achievable short-term goals, and offering information about the potential negative outcomes of not completing activities. The system also provides supplementary learning materials and exercises to support students in reaching their academic goals.
The AI system is designed to empower teaching staff to take proactive measures in addressing students’ issues.
By enabling early detection and continuous monitoring, educators can intervene before problems arise and provide timely support to students, operating on a 24/7 basis.
The research team is led by David Bañeres, who is part of the Systems, Software and Models Research Lab (SOM Research Lab) at the Internet Interdisciplinary Institute (IN3), and includes Ana Elena Guerrero, who leads the Technology-Enhanced Knowledge and Interaction Group (TEKING) and is affiliated with the Faculty of Computer Science, Multimedia, and Telecommunications. Other members are María Elena Rodríguez-González, also from the same research group and faculty, and Pau Cortadas, a researcher at the Faculty of Economics and Business.
