Low-bandwidth AI support for Chinese teaching in Tanzanian universities: A design synthesis

Artificial intelligence (AI) can expand oral practice and feedback in language learning, but its value depends on pedagogy, access, teacher capacity, assessment validity and governance. This conceptual design synthesis develops a low-bandwidth AI-supported pathway for beginner-to-intermediate Chinese teaching in Tanzanian universities. A bounded integrative review drew on 31 sources spanning AI-supported language learning, Mandarin pronunciation and speech recognition, listening-speaking assessment, instructional design, and Tanzanian digital-education policy. Evidence was interpreted through communicative can-do outcomes, Technological Pedagogical Content Knowledge, the Technology Acceptance Model, and human-centred AI principles. The synthesis produced six design requirements: pedagogy before platform, an offline-first core, targeted AI use for feedback-intensive practice, integrated listening-speaking tasks, human-in-the-loop assessment, and readiness before scale. These requirements were translated into a six-stage pathway, an oral-performance rubric, a 60-minute lesson shell, workload scenarios and an institutional readiness scorecard. The framework distinguishes direct Chinese-specific evidence from cross-language extrapolation and provides offline substitutes when connected AI is unavailable. It does not claim effectiveness: no field data were collected, and speech-recognition performance for Mandarin spoken with Tanzanian accents remains unestablished. The framework is intended for piloting, local validation and adaptation before scale.