Spatial Balancing: Harnessing Spatial Reasoning to Balance Scientific Exposition and Narrative Engagement in LLM-assisted Science Communication Writing

2025-09-17

AI Summary

This study addresses the challenge of balancing scientific rigor and narrative appeal in LLM-assisted science communication. We propose a spatialized co-writing method that externalizes revision trade-offs into a two-dimensional strategy space—scientific fidelity versus narrative engagement—where strategy labels guide the LLM to generate multiple content variants. Users navigate, compare, and iteratively refine outputs via interactive visualization. The system integrates spatial reasoning, interactive visual analytics, and large language models to enable dynamic, interpretable control over both scientific content and narrative structure. In a 16-participant controlled study, our approach significantly enhanced metacognitive reflection, writing flexibility, and creative exploration. Results validate the efficacy of spatialized representation for balanced, human-AI co-creation—marking the first integration of spatial reasoning into LLM-mediated scientific writing workflows, and establishing a novel paradigm for explainable, controllable AI-augmented authoring.

Technology Category

• Cognitive Modeling & Cognitive Systems: Computational Creativity
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Balancing scientific exposition and narrative engagement is a central challenge in science communication. To examine how to achieve balance, we conducted a formative study with four science communicators and a literature review of science communication practices, focusing on their workflows and strategies. These insights revealed how creators iteratively shift between exposition and engagement but often lack structured support. Building on this, we developed SpatialBalancing, a co-writing system that connects human spatial reasoning with the linguistic intelligence of large language models. The system visualizes revision trade-offs in a dual-axis space, where users select strategy-based labels to generate, compare, and refine versions during the revision process. This spatial externalization transforms revision into spatial navigation, enabling intentional iterations that balance scientific rigor with narrative appeal. In a within-subjects study (N=16), SpatialBalancing enhanced metacognitive reflection, flexibility, and creative exploration, demonstrating how coupling spatial reasoning with linguistic generation fosters monitoring in iterative science communication writing.

Problem

• Balancing scientific exposition and narrative engagement in communication
• Lack structured support for iterative science writing revision processes
• Integrating human spatial reasoning with LLM linguistic intelligence

Innovation

• System visualizes revision trade-offs in dual-axis space
• Users select strategy labels to generate and refine versions
• Couples spatial reasoning with LLM linguistic generation