Idea traces map how scientific ideas evolve across papers
Idea Traces: Threading Accretive Content across Scientific Papers
Human-Computer Interaction
Summary
Scientific papers often build on one another, but it can be hard to see how specific ideas change and connect over time. The authors created a method called idea traces that links content between papers to show how ideas develop and differ. They use a process with language models to break down the connections and visualize them in a way that starts from one paper and follows ideas through many others. Researchers who tried it found it helped reveal useful connections that are not easy to spot otherwise.
What this means in practice
- •For knowledge management teams: Create interactive maps of idea development across linked documents to improve knowledge discovery and synthesis.
- •For research software developers: Build tools that help researchers explore and navigate scientific ideas beyond paper-level summaries.
Authors
Luke S. Snyder, Jeffrey Heer
Abstract
We introduce idea traces, a novel thread-based representation of accretive content across scientific papers. Rather than capturing fixed themes from papers, idea traces surface similarities and deltas of faceted content that enable inspection of how ideas relate and evolve across the citation network. To extract traces, we contribute an LLM pipeline that deconstructs salient transitions between pairs of papers and collates the output into an idea dependency graph. We leverage idea traces in a visual system that supports idea-driven navigation starting from a seed paper, lifting the "plane" of synthesis from papers to ideas. To evaluate the usefulness and quality of generated ideas, we provide a case study that compares our approach with an LLM Q&A workflow, and interview 10 researchers who used our system and annotated 110 ideas. Experts found that our approach helped extract a useful space of connections that would have been difficult to make otherwise. Drawing on their feedback, we outline several promising directions for future research.