Effectiveness of adaptive versus non-adaptive intelligent tutoring systems in early primary mathematics

Adaptive intelligent tutoring systems (ITS) are considered promising tools to address increasing heterogeneity in classrooms by tailoring instruction to individual learners. However, prior research often bundles multiple adaptive features, relies on non-equivalent instructional conditions, and predominantly focuses on older learners, limiting conclusions about specific adaptive mechanisms and younger students. To address these limitations, this field experiment investigated the isolated effect of task difficulty adaptation by comparing an adaptive and a non-adaptive ITS in early primary mathematics education among second-grade students (N = 132). Both systems were identical in content, interface, feedback, and hints, differing only in the dynamic adaptation of task difficulty. We hypothesized that the adaptive system would outperform the non-adaptive system regarding students’ outcomes. Students engaged in a 30-minute training session, and pre-post assessments measured learning outcomes, cognitive load, and enjoyment. Additionally, an exploratory latent profile analysis based on motivational and cognitive learner characteristics examined differential effects. Results revealed no significant differences between conditions for learning outcomes, cognitive load, or enjoyment. Exploratory analyses indicated moderation by conscientiousness for subjective difficulty and enjoyment and differences in students' progression through the ITS. Learner profiles did not moderate the effects. These findings suggest that isolated task difficulty adaptation did not yield clear advantages over a non-adaptive system and highlight students' progression through the ITS and individual learner characteristics as promising targets for future research.