WHAT-IF: Exploring Branching Narratives by Meta-Prompting Large Language Models

2024-12-13

arXiv.org

AI Summary

This paper addresses the challenge of zero-shot generation of high-quality branching narratives from linear stories for interactive fiction (IF). Given a pre-written linear narrative, the proposed method identifies critical plot points and employs GPT-4—without fine-tuning—to generate logically coherent, stylistically consistent multi-choice branches, thereby constructing a dynamic graph-structured story representation. Its core contributions are: (1) the first plot-aware meta-prompting framework tailored for IF, enabling controllable and interpretable branch generation; and (2) a unified graph-based model for multi-path narrative states, supporting real-time interactivity and cross-branch consistency maintenance. Experimental results demonstrate that the approach preserves the original narrative’s tone while significantly improving branch plausibility, structural integrity, and player agency—achieving end-to-end interactive narrative generation.

Technology Category

• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

WHAT-IF -- Writing a Hero's Alternate Timeline through Interactive Fiction -- is a system that uses zero-shot meta-prompting to create branching narratives from a prewritten story. Played as an interactive fiction (IF) game, WHAT-IF lets the player choose between decisions that the large language model (LLM) GPT-4 generates as possible branches in the story. Starting with an existing linear plot as input, a branch is created at each key decision taken by the main character. By meta-prompting the LLM to consider the major plot points from the story, the system produces coherent and well-structured alternate storylines. WHAT-IF stores the branching plot tree in a graph which helps it to both keep track of the story for prompting and maintain the structure for the final IF system. A video demo of our system can be found here: https://youtu.be/8vBqjqtupcc.

Problem

• Creating branching narratives from linear stories using meta-prompting
• Generating coherent alternate storylines through LLM decision points
• Building interactive fiction games with structured branching plot trees

Innovation

• Uses zero-shot meta-prompting for branching narratives
• Generates alternate storylines with GPT-4 LLM
• Stores branching plot tree in graph structure