- Oct 20, 2026
- Posted by:
- Category: Abstract of 10th-educationconf
Abstract Book of the 10th International Conference on Advanced Research in Education
Year: 2026
[PDF]
AI Meets Mother Tongue: How Language Models Preserve or Erase Cultural Identity in Multilingual Academic Writing
Mohammad Sharafi
ABSTRACT:
Large Language Models and AI writing assistants have become ubiquitous tools in academic contexts, promising to improve writing quality for multilingual scholars. However, this study reveals a critical paradox: these systems systematically standardize multilingual writers’ expression toward native English norms, erasing the distinctive cultural-linguistic voices that scholars bring to global scholarship. This research compared two approaches to AI feedback with 120 multilingual graduate students (60 from Persian-speaking backgrounds, 60 from Central European backgrounds) over 8 weeks. Group A received conventional AI writing assistance providing direct corrective feedback, while Group B engaged with a human-centered AI approach generating metalinguistic scaffolding prompts. Results demonstrated significant differences: Group B achieved 58.7% transfer accuracy compared to Group A’s 28.4% (p < .001), indicating superior learning. Crucially, Group B maintained distinctive voice and identity markers in their writing (73.1% of essays) while Group A showed progressive homogenization (41.3%, χ² = 18.42, p < .001). Metacognitive awareness was significantly higher in Group B across all dimensions (p < .001). These findings challenge the assumption that more powerful AI necessarily produces better learning. AI systems designed with human-centered principles—promoting metacognitive reflection rather than automatic compliance—produce superior learning while preserving authentic multilingual voice. This has significant implications for educational technology design, language teaching pedagogy, and linguistic justice in global scholarship. The research demonstrates that technology can be ethically designed to support learner agency, cultural preservation, and genuine learning outcomes.
Keywords: Academic Writing; Artificial Intelligence; Cultural Identity; Human-Centered Design; Multilingual Learners