Educational Technologies for Multilingual Learners: A Systematic Review of AI-Based Human-Centered Design

Abstract

Background: While AI is promoted as a transformative force in education through adaptive platforms and real-time feedback, its implementation for multilingual learners often risks reinforcing educational inequalities. Current scholarship cautions that automated systems still struggle with cultural nuances and idiomatic expressions, highlighting the need for design approaches that foreground equity and human agency. Objective: This systematic review examines 10 core studies through the lens of the ISO 9241-210 Human-Centered Design framework. The objective is to analyze how AI-based educational technologies are designed and evaluated to support multilingual learners, specifically focusing on the "Context of Use," "User Requirements," "Design Solutions," and "Evaluation" phases. Methods: Through a systematic search of four databases and subsequent snowballing, 10 core papers were selected to analyze how AI tools address linguistic and cultural diversity. Results: Across the reviewed studies, reported improvements ranged from quantitative gains, including a 25–31% increase in literacy and vocabulary retention and a rise in academic success rates up to 77.8%. Furthermore, AI-driven systems, when aligned with HCD principles, were associated with saving educators up to 41% of their time. However, evaluations revealed critical 'socio-technical paradoxes': students faced a "trade-off" where they reverted to English-centric prompting due to algorithmic bias in low-resource languages, and risks of "metacognitive laziness" emerged from over-reliance on automated tools. Conclusion: The successful integration of AI in multilingual contexts depends on a shift from content generation to pedagogical scaffolding. Designers must prioritize "Human-in-the-loop" models that balance computational efficiency with human oversight to provide the emotional engagement and cultural authenticity that AI currently lacks.