Learner Behavior Profiles during Problem Formulation in Problem-based Learning

Abstract

This study examines student learner behavioral patterns during problem formulation in a Problem-Based Learning (PBL) activity conducted in a micro-learning platform with integrated generative AI support named LA-ReflecT. Students’ interaction log data were analyzed to understand how different behavioral strategies relate to the quality of the generated problems. Interaction sequence data from 70 students were modeled using the Longest Common Subsequence (LCSS) similarity measure and clustered using the K-Medoids algorithm. Three distinct behavioral profiles emerged: Focused and goal-oriented, Exploratory and inconsistent, and Balanced help-seeking. Although students exhibiting focused and balanced behaviors achieved slightly higher average problem-quality scores, there were no statistically significant performance differences across the three profile clusters.