Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208350,
        author = {Jamilu Auwalu Adamu and Muhammad Bello Sadiq and Professor Promise Mebine},
        title = {MathemAfrica: A Multilingual Explainable Artificial Intelligence Ecosystem for Adaptive Mathematics Education in Resource-Constrained African Learning Environments},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {4},
        pages = {3219-3233},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208350},
        abstract = {Secondary mathematics education in developing and resource-constrained contexts confronts systemic bottlenecks, including overcrowded classrooms, acute teacher shortages, frequent electrical blackouts, prohibitively expensive cellular data, and linguistic dissonance between official English instruction and students' indigenous home languages. To address these interconnected challenges, this study developed and evaluated MathemAfrica, an innovative digital learning platform integrating Bayesian Knowledge Tracing, Multidimensional Item Response Theory (M-3PL IRT), offline optical handwriting verification, and multilingual scaffolding in Hausa, Yoruba, Igbo, and Swahili. Employing an ADDIE-based design science methodology and a 16-week quasi-experimental non-equivalent pretest-posttest control group design, the intervention was evaluated across six public secondary schools in Kano State, Nigeria, with 1,428 Grade 10 students. Expert reviews confirmed high content validity (97.06%) and media quality (94.56%), while students reported exceptional practicality (96.13%). Analysis of Covariance revealed statistically significant and substantial gains in mathematical proficiency, with adjusted posttest scores of 73.84% in the intervention group versus 57.22% in conventional instruction (F (1, 1425) = 148.92, p < .0001, Cohen's d = 0.94). Furthermore, the platform closed 80.4% of the indigenous language achievement penalty, while maintaining real-time adaptive sequencing at merely 38.4 kB per session. 
The results provide empirical evidence that combining edge-native computation with culturally responsive multilingual scaffolding strengthens secondary mathematical achievement under severe resource constraints.},
        keywords = {Adaptive learning, Artificial intelligence, Bayesian Knowledge Tracing, Educational equity, Explainable artificial intelligence (XAI), Item Response Theory, Mathematics education, Multilingual scaffolding, Resource-constrained environments.},
        month = {September},
        }