In this project, we developed a 2D game based on Chinese ancient painting The Night Revels of Han Xizai (Ming Dynasty replica.) We employed LLM-driven npc and LLM-driven clue system to empower the gamification. The players can flexibly travel between scenes in the painting and talk with different characters in the scene, diging out information and exploring the stories behind the painting. As they go deeper, they will collect clues about the painting, which is controlled by the LLM-driven systems. The clues are designed to provide cultural backgrounds and stories, and we planned to carry out studies to exam the trustworthiness and credibility of the conversation, clues and overall experience provided by the LLM-system we designed.
1.How do translators interact with LLM-based multi-agent systems to simulate the cultural heritage narratives in a trustworthy way?
2.How do users identify and evaluate sources of trust in their interactions with LLM-based and without LLM-based multi-agent cultural heritage narrative systems?
Diagram of technology framework.