JMIR Publications

ABSTRACT

Generative AI is arriving in high‑stakes assessment, yet governance, validity evidence, and faculty readiness remain uneven. From a Taiwan–Japan perspective, we outline a pragmatic, transferable approach to integrating AI into nursing OSCE using a Five AI Roles model—Learning Assistant, AI‑augmented Standardized Patient (AI‑SP), Assessment Assistant, Case Generator, and Learning Analyst—mapped across pre‑, peri‑, and post‑OSCE workflows with human‑in‑the‑loop final judgment. Taiwan contributes agile edu–engineering co‑development, staged pilots (practice → mock OSCE → limited high‑stakes stations), A/B comparisons, and explainability‑by‑design logging that links scores to time‑stamped evidence. Japan contributes robust policy scaffolding (national AI use guidance in K–12, a revised nursing Model Core Curriculum with outcomes and assessment blueprints, and institutional research cultures that support auditability and quality assurance). We distill four cross‑cutting governance pillars—human oversight, learning‑process transparency, ethics and safety, and traceability—into implementable techniques (machine‑readable rubrics, SP persona cards, bias monitoring, and targeted faculty development). Aligning with international principles (IACAI, OECD, UNESCO/WHO, EU HLEG, NIST), we propose a joint roadmap and shared registry to benchmark reliability, validity, equity, and workload impact. A Taiwan‑led agility complemented by Japan’s standards‑driven assurance can form an Asia–Pacific reference model for trustworthy AI‑augmented OSCE in nursing education.