The Effectiveness of Artificial Intelligence-Assisted Problem-Based Learning Model to Improve Elementary School Students' Mathematical Representation Processing Ability

Mathematical representation processing ability constitutes a foundational cognitive competency for elementary students, yet persistent deficits in this domain across Indonesian schools call for innovative, evidence-based instructional solutions. This study examines the effectiveness of an Artificial Intelligence-Assisted Problem-Based Learning (AI-PBL) model in improving mathematical representation processing ability among fifth-grade elementary school students in Lampung Province. Employing a quasi-experimental Non-Equivalent Control Group Design, 60 students from a single elementary school in Lampung were assigned to an experimental class (n=30, receiving AI-PBL) and a control class (n=30, receiving conventional instruction). Data were collected via a validated 10-item representation test and analyzed using the Shapiro-Wilk normality test, Levene's homogeneity test, independent samples t-test, normalized N-Gain, and Cohen's d effect size. The experimental class achieved a posttest mean of 80.07 (pretest: 52.43), while the control class reached 64.90 (pretest: 52.23). The independent t-test yielded t=6.317 (p<0.001), confirming a statistically significant difference. Mean N-Gain was 0.597 (moderate) for the experimental class versus 0.275 (low) for the control, with Cohen's d=1.631 indicating a very large effect. These findings confirm that AI-PBL is highly effective in developing elementary students' mathematical representation processing ability, contributing a replicable, AI-integrated pedagogical model for primary mathematics education.