Co-Designing Curricula with Artificial Intelligence: A Hermeneutic Phenomenological Inquiry into Primary School Teachers' Agency and Pedagogical Negotiation in Indonesia's Kurikulum Merdeka

Background: The unprecedented diffusion of generative artificial intelligence (GenAI)—predominantly large language models (LLMs) such as OpenAI’s ChatGPT—is fundamentally restructuring instructional planning and curriculum design across global basic education. In the Global South, exemplified by Indonesia's nationwide rollout of Kurikulum Merdeka (the Emancipated Curriculum), primary school educators face intensifying administrative demands to independently engineer differentiated, contextualized instructional modules (Modul Ajar). While existing literature predominantly evaluates technology adoption through instrumental, quantitative survey frameworks (e.g., TAM, UTAUT), the subjective, pre-reflective, and existential lifeworlds of primary teachers co-designing curricula alongside AI algorithms remain critically under-theorized. Objectives: This study investigates the lived experiences (Erlebnis) and meaning-making processes of primary school teachers integrating ChatGPT into their instructional design workflows, examining how algorithmic interaction reshapes teacher agency, professional identity, and pedagogical curation. Methods: Utilizing an interpretive hermeneutic phenomenological design grounded in Max van Manen’s Phenomenology of Practice and conceptualized through Priestley, Biesta, and Robinson’s Ecological Model of Teacher Agency, this inquiry engaged certified classroom educators at SD Negeri 193 Pekanbaru, Indonesia. Qualitative data were generated over a four-month period through iterative in-depth semi-structured phenomenological interviews, non-participant planning observations, and curricular textual audits comparing raw ChatGPT conversation logs directly against finalized, officially approved Modul Ajar documents. Data analysis followed Moustakas's horizontalization and van Manen's thematic hermeneutic isolation in strict adherence to COREQ guidelines. Findings: The phenomenological reduction revealed four invariant thematic structures: (1) Epistemic Scaffolding and Curricular De-constriction, wherein ChatGPT acts as a cognitive sounding board that alleviates the paralysis of the blank page and compresses manual retrieval time; (2) The Imperative of Critical Pedagogical Curation, embodying a rigorous human-in-the-loop governance stance that actively detects, edits, and purges algorithmic hallucinations, developmental mismatches, and socio-geographic alienation; (3) The Dialectic of Prompt Craftsmanship and Material Negotiation, illustrating that machine utility is an emergent manifestation of teacher pedagogical intentionality tempered by the structural constraints of free-tier technology; and (4) The Inviolable Sanctuary of Affective and Moral Pedagogy, demonstrating that pastoral care, emotional attunement, and character formation (Profil Pelajar Pancasila) represent uniquely human, embodied vocational prerogatives that algorithms can never replicate. Conclusions & Significance: Far from automating educators into obsolescence, generative AI catalyzes relational teacher agency within a socio-material co-design ecology. Planning labor is fundamentally elevated from clerical transcription to high-level critical curation. We propose actionable frameworks for integrating critical AI prompt literacy into teacher education institutes (LPTK) and formulating national ethical policies for AI-assisted curriculum development.