Tracing Ecological Metaphors in Discourses on Open Science using LLMs and Knowledge Graphs

The term "ecosystem" is frequently used to describe various concepts, not only in open science but also in broader discussions of research and innovation. Despite its widespread use, it is rarely explicitly defined, often functioning as a boundary object that facilitates communication across diverse communities. Systematically documenting its varied, context-dependent meanings presents a significant challenge. This work in progress explores the term "ecosystem" within the discourse on open science, offering a systematic approach to mapping its varied meanings and uses. We pose a twofold research question: from a social scientific perspective, how can the diverse uses of "ecosystem" be systematically documented? And from a methodological standpoint, how can computational techniques be leveraged to trace such a boundary object? Drawing on a curated corpus of 211 scholarly articles and exploratory ontological work, we use LLMs to construct a detailed knowledge graph, yielding 1,067 semantic relations. This graph is then integrated with a citation network to create a multilayer model for analyzing the term’s dissemination. Our preliminary results identify seven distinct, data-driven thematic communities. Although the application of knowledge graphs is now an emerging practice, our pipeline offers a novel application for revealing the term’s underlying meanings. By mapping its surrounding ontology, this ongoing work suggests how such a term allows knowledge to circulate between different scholarly communities, providing deeper insight into the conceptual landscape shaping the digital transition of research.