Negotiating Educational Authority in Human-AI Collaboration: Exploring Chinese Language Teachers’ Agency in AI-Assisted Task Design

Abstract

Generative AI (GenAI) can be used to create lessons and tasks, but its pedagogical effectiveness depends on teachers’ professional judgments about whether to accept, modify, or reject AI suggestions. This qualitative multiple-case study investigates the professional agency of six Chinese language teachers who created or revised beginner-level shopping tasks with ChatGPT. Data comprised semi-structured interviews, time-stamped ChatGPT interactions, and teacher revisions, allowing comparison between stated beliefs and observable design decisions. Reflexive thematic analysis and cross-case comparison identified three interrelated patterns. First, teachers filtered AI recommendations through existing pedagogical knowledge and beliefs, including judgments about learner level, syllabus content, instructional sequencing, and task-based language teaching (TBLT). Second, teachers localized AI-generated tasks to learners’ interests and classroom ecology, especially by reducing cultural distance and simplifying classroom-management demands. Third, efficiency shaped uptake: some teachers were highly selective and retained little AI-generated content, whereas others prioritized efficiency and accepted most of it. The findings conceptualize AI-assisted task design as a negotiation of pedagogical authority rather than simple technology adoption. GenAI expanded teachers’ options and reduced preparation demands, but teachers still had to ensure contextual, theoretical, linguistic, and practical fit. The study contributes process-level evidence to emerging research on language-teacher agency and GenAI and to technology-mediated TBLT by showing how task-design principles are applied, adapted, or neglected in human-AI interaction.