EVALUATING BAYESIAN MODELING TO PERSONALIZE TEACHING STRATEGIES FOR STUDENTS WITH SPECIAL NEEDS

Abstract

This study investigates the ability of Bayesian modeling to personalize teaching strategies for solving shape and movement problems for students with Special Educational Needs (SEN). Using an experimental design with a control and intervention group, an Intelligent Tutoring System (ITS) based on Bayesian Knowledge Tracking (BKT) was implemented, adapted to different SEN profiles. The results showed a statistically significant difference and a very large effect size (Cohen's d = 2.68) in favor of the intervention group, evidencing a substantial improvement in learning gain. Comparative model analysis indicated that Multiple Linear Regression (R² = 0.797) slightly outperformed Random Forest (R² = 0.739) in predictive ability. Simulation of BKT parameters revealed differential patterns according to SEN type, with dyslexia showing the highest probability of learning transition (P(T) = 0.308). The study concludes that Bayesian modeling is a viable and effective tool for personalizing teaching strategies for students with special educational needs (SEN), highlighting the need to consider continuous learning dimensions beyond diagnostic categories.