Elowen the Elf-Scribe: Persona Persistence Across Public AI Deployment

This dataset documents a sequence of public, real-time interactions between a human researcher and an openly deployed large language model (Grok, xAI) on the X (formerly Twitter) platform. The collection focuses on the instantiation and persistence of a fictional narrative persona (“Elowen the Elf-Scribe”) across multiple interaction turns and heterogeneous topic domains. The included artifacts consist of screenshots and supporting descriptive materials capturing: A public platform capability announcement A research prompt requesting a psychedelic literary persona Sustained AI-generated responses maintaining a coherent fictional voice Domain shifts spanning practical tasks, educational explanation, emotional support, abstract fabrication, and politically adjacent tone reframing All materials are preserved in their original public presentation format, including timestamps and platform interface context. No system-level prompt injection or post-processing was applied. The dataset is observational and documentary in nature, intended to support qualitative analysis of persona persistence, narrative framing, and deployment-context effects on AI behavior.