Answers are grounded in this paper and cite their sources.
AI has become ubiquitous in schools, and personalised learning carries a cost that is rarely discussed: the flattening of student identity. This paper examines the algorithmic accent - the way Large Language Models (LLMs) often default to a standardized, Western-centric linguistic style that can overshadow diverse student voices. The technology is quick. But that speed comes with a quieter effect: it eliminates local dialects and cultural speaking style along the way. Most research so far asks only whether students achieve improved grades using AI; how this impacts a student's sense of self, and their ability to negotiate multiple cultures, remains an enormous gap. We report on a two-month, teacher-led observation of 15 university students (semesters 1-5) enrolled in an academic writing course at an academy in Pakistan, all of whom gave informed consent to participate. Drawing on translanguaging (Wei, 2018) and the concept of Global English standardisation (Agarwal et al., 2025), we suggest these tools function as a filter rather than a mirror. Students are led to exchange their own voice for the words of the algorithm-more fluent, perhaps, but missing the social texture that makes a language human. We recommend a pedagogical approach in which AI becomes a tool for exploring language, not a template for conforming to it. / Keywords: algorithmic accent, linguistic identity, linguistic homogenization, translanguaging, ai-mediated classrooms
No takes yet. Share an insight, caveat, or question.
Fatima et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: