<![CDATA[English language evaluation in West Sulawesi has long relied on culturally foreign materials, producing a digital cultural gap that alienates students from Mandar identity, while an emerging AI-centric paradigm risks eroding teachers' pedagogical agency. This study examines the dialectic and negotiation of meaning undertaken by five English teachers in the Mandar region as they transformed local ethnoliteracy including the sandeq boat, the Metawe' politeness tradition, and the Boyang traditional house into interactive English evaluation instruments through a Human-AI Co-Creation strategy. Adopting a qualitative, document-based Participatory Design, this article reports findings from reflexive thematic analysis of five prompt-engineering logs (coded T1–T5) produced during a Human-AI Co-Creation Workshop involving four generative AI platforms (DeepSeek, ChatGPT, Claude, and Gemini). Analysis identified five interrelated themes: contextualizing prompts through curriculum and local culture, negotiating completeness of the evaluation structure, negotiating linguistic-level appropriateness, technical dialectic over AI system limitations, and validating substance to filter cultural hallucination. These themes were synthesized into a preliminary five-stage model of teacher–AI dialectic: Contextual Initiation, Initial AI Generation, Critical Teacher Audit, Negotiation and Iterative Revision, and Resolution. The findings indicate that teachers function as active co-designers rather than passive operators, and that the cultural accuracy of AI output depends heavily on the richness of the contextual input teachers supply at the outset. This document-analysis phase forms part of a broader study; teachers' phenomenological experience of the dialectic process will be examined separately through hermeneutic-phenomenological interview data in a subsequent publication.]]>
Copyrights © 2026
