Re-experiencing history: a platform for the re-enactment of historical events with multimodal large language models

This paper presents "Re-Experiencing History," an innovative project leveraging AI-powered large language models (LLMs) and image generation technologies to reconstruct historical events. Inspired by criminal investigation techniques, the project aims to create a platform enabling historians, students, and the public to visualise and engage with historical scenes. The pilot study will focus on two ancient Roman scenarios: the triumph and the Lupercalia festival. Using DALL-E 3, we have already generated 600 images based on carefully crafted prompts derived from historical literature. Initial evaluations reveal varying degrees of prompt-to-image alignment, highlighting potential knowledge gaps. The project's interdisciplinary approach combines computational linguistics with historical expertise to enhance the accuracy and utility of AI-generated historical recreations. By bridging past events with modern technology, this research contributes to digital humanities and explores new methodologies for historical study and education. Files