Biranchi Poudyal

Introduces Institutional AI Governance Stress Testing (IAGST), a method combining controlled capability escalation, a frozen documentary corpus and a six-dimensional governance response chain to locate where university AI governance breaks down. Applied across 133 documents from five Western Australian universities using 15 scenarios (75 encounters), the results show that while responsible authority roles typically existed, they lacked sufficient decision criteria, and that governance reliability declined sharply as AI use moved from augmentation toward autonomous substitution.

Examines how generative AI redistributes epistemological authority in education — from teachers, to students, to machines. Reviewing this shift through Distributed Agency, Self-Determination Theory, TPACK and SAMR, the paper argues existing frameworks fail to address equitable distribution of power, data ownership and accountability, and introduces the Ecological Co-Agency Framework, which anchors relational, regulatory and pedagogical dimensions of agency in human accountability for knowledge production.