The effectiveness of AI-based emotional feedback in an online mathematics game on fractions for junior high school students

Mathematics learning often triggers anxiety and motivation loss when students encounter persistent failure. The present study examines the impact of an online math game featuring AI-based emotional feedback on students’ learning outcomes in fraction topics. Specifically, it analyzes the effectiveness of real-time AI emotional feedback in improving academic performance among junior high school students. A sequential explanatory mixed-methods design was employed. The study involved 119 eighth-grade students aged 13–14 years, with balanced gender representation across two research sites. Participants were assigned to an intervention group (n=61) receiving AI feedback and a control group (n=58). Quantitative analysis using t-tests revealed that the intervention group significantly outperformed the control group in both School X (t = -5.43, p < .001, d = 1.42) and School Y (t = 2.22, p < .033, d = 0.57), indicating medium-to-large effect sizes. System log data showed that “Neutral” was the most frequent emotional state, while “Sadness” emerged as the primary negative emotion during challenging tasks. Qualitative findings further indicated that AI feedback supported cognitive reappraisal and refocusing strategies, enabling students to transform negative emotions into persistence. Overall, AI-based emotional feedback functions as effective affective scaffolding, mitigating the impact of negative emotions on mathematical performance.