A possible prompt structure for AI director in narrative games

As narrative designers, we always face the dilemma between linear narrative and infinite branches for designers to make the story more diverse, which means we want the players to have personalized narrative experience, but are afraid of the workload and the possibility that players will have too much freedom to enjoy the intended story flow we designed. But with AI advancing, more and more AI applications are appearing. As for large language models (LLMs), the ability of natural language understanding (NLU) and natural language generation (NLG) they have aligns well with creating story, which suggests that AI can be the key to this problem, helping us go further to the aim of Holodeck (Murray, 1997). In this research, I want to explore a possible prompt structure that can enable AI to act as a director, dynamically generating stories based on player actions while following designers’ intention.