Reading the paper, it appears only 7 people used the LLM version, and the LLM had a detailed prompt that had several paragraphs handwritten for the canned module that the LLM was running, and of the 7 players, 3 reported feeling railroaded. Further, the scores for creativity were based on the players asking for a room description and the resulting text was what was scored for creativity. Additionally, the system was used for a one-player only game without social interaction.A friend notified me of this: https://boingboing.net/2025/01/14/ai-might-be-your-next-dungeon-master.html