NarrativeSteward helps authors manage AI-generated interactive stories
NarrativeSteward: Coordinating Delegation, Guidance, and Verification in Agent-Assisted Interactive Narrative Authoring
Human-Computer InteractionArtificial IntelligenceComputation and Language
Summary
Creating interactive stories with AI agents can produce a lot of content, making it hard for authors to keep track. The authors developed NarrativeSteward, a tool that organizes story elements and helps writers understand and guide the AI's work. It lets authors review the story structure and changes, making it easier to request edits and verify results. Tests and a study showed it improved authors’ ability to manage and revise interactive narratives compared to general tools.
What this means in practice
- •For interactive game designers: Manage and refine AI-generated storylines with linked outlines and verification tools integrated in authoring environments.
- •For software developers in creative tools: Build user interfaces that improve author control over AI-assisted content creation and revision through structural reviews and change tracking.
Authors
Wenjin Wang, Jiazhen Lei, Yuxin Sha, Nuwa Xi, Meng Zhao, Xingxi Yin, Qi Liu, Yuliang Shen, Zixun Sun
Abstract
Autonomous AI agents can turn authors' goals into interactive narratives by independently organizing and carrying out generation and revision. As agents generate and revise extensive content, authors struggle to grasp its overall structure, local details, and relationships, complicating continued guidance. We present NarrativeSteward, an authoring environment that organizes outlines, worldbuilding, and narrative graphs as linked artifacts for agent implementation and author guidance. Agent dialogue and project-wide structural review help authors understand the evolving work and guide local and cross-layer revisions, while change records and execution verification help authors assess the resulting work. Technical tests validated the system's change records, recovery mechanisms, and execution diagnostics. In a 12-participant within-subject study, NarrativeSteward supported easier formulation of revision requests and inspection of changes, and greater perceived understanding of changes and story structure, than general-purpose agents. Qualitative findings show how reviewing the work and feedback helps authors develop requirements and guide subsequent delegation. We open-source NarrativeSteward at https://github.com/Tencent/NarrativeSteward.