Answers are grounded in this paper and cite their sources.
Science fiction serves as a powerful medium for exploring future technologies and their societal impact, offering rich opportunities for entertainment experiences that engage audiences with speculative futures. However, existing AI-based story generation approaches often lack systematic grounding in social-cultural analysis, limiting their potential for creating meaningful interactive entertainment content. This paper presented a novel multi-agent system that combines the Archaeological Prototyping (AP) methodology with large language model collaboration to generate science fiction stories with enhanced diversity while maintaining quality. Our approach operated in two phases: first, we constructed an AP model through a three-stage pipeline using evidence-based modeling and multi-agent competitive prediction; second, we employed a hierarchical multi-agent architecture for narrative generation, where an Overseer Agent coordinated Setting and Outline Agents while maintaining consistency with the AP model. Experimental results demonstrated that compared with the baseline method, our multi-agent approach improved story diversity while maintaining quality—a trade-off that temperature-based approaches fail to achieve. The generated narratives can serve as foundations for various entertainment applications, including interactive fiction, games, and trans-media storytelling experiences. Lastly, this architectural pattern, combining deterministic structural grounding with multi-agent ideation, is transferable beyond the AP model to other domain-specific ontologies and structured knowledge frameworks.
No takes yet. Share an insight, caveat, or question.
Menghan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: