Using Clustering Techniques to Design Learner Personas for GenAI Prompt Engineering and Adaptive Interventions
Ivan Tudor, Martina Holenko Dlab, Gordan Đurović, Marko Horvat · Electronics · 2025
Personalized learning in higher education aims to enhance student motivation, engagement, and academic outcomes. Learner personas as representations of students offer a promising approach to personalizing learning in technology-enhanced environments, particularly in combination with learning analytics (LA). This study explores how LA can be used to identify activity patterns based on data from the E-Learning Activities Recommender System (ELARS). The activity data of STEM students (N = 90) were analyzed using K-Means clustering. The analyses were based on timing, the percentage of task completion, and their combination to identify distinct engagement patterns. Based on these, six clusters (learner personas) were identified: consistent performers, overachievers, last-minute underperformers, low-engagement students, late moderate achievers, and early proactive performers. For each persona, GenAI prompts and personalized interventions based on motivational and instructional frameworks were proposed. These will inform further development of the ELARS system, with the goal of enabling personalization, promoting self-regulated learning, and encouraging students to integrate GenAI tools into their learning. The study shows how the combination of clustering techniques for learner persona development with GenAI prompt engineering and adaptive interventions has the potential to drive the design of personalized learning environments.