How Many Kilometers Can I Drive Before Shisa-chan Gets Sick? Affective LLM-AR Gamification for Mobility Behavior Change in Okinawa

Reducing CO2 emissions is often framed as a collective good, yet it remains a distant and abstract concept in the context of everyday life. This study seeks to translate these remote and hard-to-quantify emission reduction goals into easily comprehensible pet-nurturing actions. By gamifying affective interactions mediated through a mobile-based augmented reality (AR) agent powered by a large language model (LLM), we developed an embodied representation of Shisa, the guardian spirit of Okinawa Island. Through explicit representation of the island’s transportation-related emissions using Shisa’s emotions and advice, our preliminary study with stakeholders, including local residents and government representatives, revealed increased emission awareness (e.g., feelings of guilt) and social tensions that emerged when participants considered changing their transport modes (e.g., negotiation). We conclude with a reflection on the practical and societal impacts of our study and future research plans for large-scale empirical deployment.