#1262: Learning from the Experts: Exploring Human–AI Feedback through Expert-AI

This demo introduces Expert-AI, a Retrieval-Augmented Generation (RAG) system designed to emulate aspects of human experts’ reasoning and feedback. Drawing on experts’ writings, lectures, and interviews, the system provides dialogic, context-sensitive feedback grounded in expert perspectives, allowing experts to review and refine responses over time. We report preliminary findings from a pilot study in a business idea creation course at a private university in Japan, where students used both Expert-AI and general-purpose generative AI systems (e.g., ChatGPT). The findings suggest complementary roles: students used Expert-AI for conceptual framing and strategic evaluation, and general-purpose generative AI for elaboration and implementation planning. These findings suggest learners’ interpretations of AI feedback may vary depending on how it is presented. Accordingly, the demo presents the text version of Expert-AI alongside early multimodal prototypes including voice and android embodiments to examine how interaction modality shapes perceived credibility, empathy, trust, and engagement in expert-like AI feedback.