Nunchi-Aware Multi-Agent Chatbots for Socially Contextual Response Generation

This paper presents a novel approach to integrating the Korean cultural concept of "Nunchi," which refers to the ability to perceive social cues and respond appropriately, into chatbot systems. Nunchi plays a central role in maintaining harmony within collectivist societies, yet it remains largely overlooked in artificial intelligence design. We propose a lightweight framework that embeds the Nunchi Scale, a validated psychological instrument, into a multi-agent chatbot platform without requiring extensive data training. The system dynamically processes conversations using a dual-memory structure that reflects both Nunchi Perception and Nunchi Action. A user study involving 130 Korean participants evaluated the system’s contextual sensitivity. Participants consistently identified the Nunchi-enabled chatbot in multi-party dialogues, confirming the effectiveness of the cultural and emotional adaptation. This research demonstrates the feasibility of culturally intelligent AI and offers a pathway toward more human-like communication through the integration of emotion, context, and cultural norms. The proposed framework is efficient, generalizable, and applicable to a wide range of socially intelligent AI systems.