Speaker
Description
Abstract: Generative artificial intelligence can now produce well-organized, linguistically sophisticated responses to many conventional university speaking prompts, complicating teachers’ efforts to determine whether a prepared oral performance represents learners’ own reasoning, language production, and communicative competence. This conceptual and pedagogical study develops a framework for designing AI-resilient speaking tasks in university English classrooms. It draws on a targeted literature review, document analysis, deductive thematic analysis, conceptual analysis, and framework-based task design. The analysis identifies six recurring vulnerabilities in conventional speaking tasks and reformulates them as six complementary design principles: contextual specificity, learner ownership, process visibility, interactional contingency, judgment and justification, and transparent AI engagement. These principles are operationalized through a five-step design sequence that begins with specifying the intended communicative evidence and appropriate AI condition, then situates learners in a meaningful context, requires a defensible judgment, introduces live adaptation, and makes preparation and AI use partially visible. The framework is illustrated through problem-solving discussions, changing-condition role-plays, AI-output critique, evidence-based mini-presentations, and dialogic position tasks. Because the framework is conceptual, its feasibility and validity require empirical examination. AI resilience does not mean making tasks entirely immune to AI; it means preserving credible evidence of what students can understand, decide, explain, and negotiate when AI is available.
