Toward Intelligent Agent-Based Scientific Storytelling

Introduction

Over the past few decades, a significant quantity of scientific papers have been published. However, many of them remain unread, despite their potential level of novelty. Traditional dissemination formats, based on static papers or technical reports, are often inaccessible to policymakers, executives, or the general public. This paper proposes a vision for an intelligent framework that utilizes Generative AI agents to transform scientific papers into structured, audience-aware narratives, leveraging storytelling principles to enhance comprehension and engagement.

Methods

The proposed methodology introduces a multi-agent pipeline with three components. A Narrative Planner Agent first analyzes the paper and constructs a storyline using a three-act structure (Context, Problem, and Solution) [1]. A Generative Presenter Agent then produces a presentation tailored to the target audience. Finally, a Rubric-Based Evaluator Agent assesses the generated presentation on criteria such as clarity, scientific rigor, and audience alignment. The entire process is envisioned within an open-source workflow automation environment (e.g., n8n [2]) to ensure scalability and modularity.

Results

At this stage, the methodology is in the design phase. However, preliminary texts in other domains have demonstrated that Generative AI can produce audience-based stories originating from domain-specific material [3].

Conclusions

This work outlines a path toward AI-driven scientific storytelling, where research is transformed into meaningful, contextualized narratives. Such systems could pave the way towards interdisciplinary communication, science education, and public engagement in applied sciences.