Artificial Intelligence and Digital Platforms in Language Education at Bangladeshi Higher Education Institutions: A Critical Narrative Review of Evidence, Mechanisms and Structural Constraints
Sharmin Rahman Bipasha *
Department of English, CCN University of Science & Technology, Cumilla, Bangladesh.
*Author to whom correspondence should be addressed.
Abstract
Generative artificial intelligence and institutional digital platforms have entered university language classrooms in Bangladesh faster than the research base needed to evaluate them. This article critically reviews the evidence produced within and about Bangladeshi higher education, reading it against the wider international literature on computer-assisted language learning, automated feedback and generative writing tools. Sources were identified through structured searching of openly accessible scholarly indexes and citation databases, supplemented by backward and forward citation tracing and by examination of authoritative institutional documents, with the final search conducted on 5 June 2026. The synthesis is organised around five analytical problems: the structural and sociolinguistic conditions that precede adoption; the epistemic limits of the perception and acceptance studies that dominate the national corpus; the contested evidence on writing and feedback; the near-absence of evidence on speaking, interaction and multilingual practice; and the governance vacuum surrounding academic integrity. Three conclusions are reasonably secure. Adoption among Bangladeshi students and teachers is already extensive and largely informal; teacher preparedness is constrained by institutional rather than individual factors; and no study conducted in Bangladesh has yet measured language learning outcomes under controlled conditions. Claims about improved writing, speaking or engagement in this setting therefore rest almost entirely on self-report. The international experimental literature on which such claims are implicitly modelled reports effects that are heterogeneous, sensitive to learner digital literacy, and accompanied by evidence of cognitive offloading and reduced authorial control. The review argues that the prevailing framing of artificial intelligence as an access-equalising technology is inadequately supported in a system already stratified by uneven connectivity, examination-oriented assessment and contested medium-of-instruction policy. Priorities for subsequent work include outcome-based designs anchored in Bangladeshi curricula, investigation of tool performance on Bangla-English bilingual production, and empirical study of institutional policy formation rather than its absence.
Keywords: Artificial intelligence in education, English language teaching, Bangladesh, higher education, generative artificial intelligence, automated writing evaluation, educational technology policy, digital divide