Vision-language agents for sustainable tourism
摘要
This study develops and evaluates an image-grounded sustainable tourism guidance app designed to promote sustainable tourism. The application utilizes vision-language models and AI agents to transform minimally annotated destination images into actionable sustainable travel advice. By fine-tuning models on destination-specific data, the system ensures accurate location recognition, which is essential for providing grounded environmental, cultural, and social recommendations. Phase 1 demonstrates that fine-tuning on Liechtenstein-specific imagery improved location recognition from 40 to 95%. Phase 2, involving 204 tourists, utilized PLS-SEM to confirm that attitude, subjective norms, and PBC significantly drive app usage intention, with environmental concern amplifying these effects. These findings offer theoretical insights for IS and tourism research while providing practical guidance for destination management organizations seeking to deploy image-grounded sustainable tourism guidance apps.