A Comparison of the Effectiveness of AI-Assisted Learning versus Conventional Media for Children

This study examined the comparative effectiveness of artificial intelligence (AI)-assisted learning and conventional instructional media among fourth- and fifth-grade elementary school students. Although AI is increasingly promoted as a means of personalizing instruction, empirical evidence directly comparing AI-based and conventional approaches at the elementary level remains limited. A quasi-experimental Pretest-Posttest Control Group Design was employed, involving 42 students in an AI-assisted experimental group and 42 students in a conventionally taught control group drawn from intact classrooms within the same school. Learning effectiveness was operationalized through normalized gain (N-Gain) scores derived from pretest and posttest achievement tests previously established as valid and reliable. Because the experimental group's N-Gain scores departed from a normal distribution, between-group differences were examined using the Mann-Whitney U test, complemented by Cohen's d to estimate practical significance. Results showed that although the two groups began with statistically equivalent pretest scores, the experimental group achieved a significantly higher N-Gain (M = 0.40, medium category) than the control group (M = 0.26, low category), U = 1190.500, p = .006, with a medium effect size (d = 0.551). These findings indicate that AI-assisted learning produced greater and more consistent improvement than conventional media among elementary learners. The study contributes empirical evidence to the underrepresented area of AI-supported elementary education and offers practical guidance for teachers, curriculum developers, and policymakers integrating AI into classroom instruction.