AI-Driven Virtual Patient Simulation for Midwifery Education

• Gael Lucero-Palacios, Computer Science, University of Delaware

• Leila Barmaki, Computer & Information Sciences, University of Delaware

While technology has improved many aspects of healthcare, there remains a notable gap in hands-on educational training for aspiring healthcare professionals, particularly in midwifery. Our project innovates this learning space by immersing students in simulated clinical scenarios with a virtual pregnant patient. Previously, these simulations relied on a “human-in-the-loop” design, requiring an instructor to manually control the avatar’s dialogue, which often resulted in restricted, scripted conversations. Because these rigid scripts cannot easily adapt to the unpredictable nature of real human interaction, students often miss the critical opportunity to navigate sudden emotional shifts. To overcome these limitations, our current work integrates artificial intelligence to fully automate the virtual patient. This AI-driven system eliminates the need for a human controller, allowing nursing students to practice independently. Furthermore, the AI enables highly natural, dynamic interactions, moving beyond surface-level questions to ask follow-up questions, request clarification on complex medical terminology, and realistically express patient concerns regarding sexual health, childbirth, and postpartum care. We hypothesize that transitioning to this AI-powered simulation will provide a significantly more authentic, scalable, and accessible training environment. Ultimately, this autonomous approach aims to enhance students’ conversational skills, confidence, and professional composure when navigating sensitive or unexpected medical dialogues in their future clinical careers.