Chasing Certainty About Uncertainty: My Journey on InSciM
Over the past several years I’ve had the chance to work on the ANR InSciM project, an interdisciplinary effort to model how uncertainty is expressed in scientific writing. I joined first as a PhD Fellow and have since continued with the team as a Postdoctoral Researcher, based at CRIT, University of Franche-Comté.
What InSciM is about
Science is full of uncertainty — margins of error, inductive reasoning, methodological limits — and how researchers express that uncertainty in their writing varies a lot across disciplines. InSciM sets out to build a linguistic model of scientific uncertainty that works across fields, from the Social Sciences and Humanities to Science, Technology, and Medicine, and to turn that model into annotated datasets and tools that can detect and categorize uncertainty automatically.
What I worked on as a PhD Fellow
During my PhD, my focus was on the core annotation framework at the heart of the project: defining what “uncertainty” looks like at the sentence level, and building a multi-dimensional scheme that could be applied consistently across very different kinds of scientific texts. That work fed directly into the Gold Standard annotated datasets the project has released, and into comparing knowledge-based (pattern-based) approaches against machine learning models for detecting uncertainty automatically. I was also involved in mentoring newer members of the team as the project grew.
What I worked on as a Postdoctoral Researcher
As a postdoc, I’ve moved from building the annotation framework to extending and applying it — working with new interdisciplinary corpora, refining the datasets for broader use by the research community, and exploring how these tools can support real semantic search and exploration interfaces for uncertainty in scientific literature.
Why this work matters to me
What keeps me engaged with this project is the bridge it builds between NLP methods and a genuinely interdisciplinary question: how confident should we — and should the public — be in what science currently says? Better tools for surfacing uncertainty in scientific texts feel like a small but concrete contribution to more honest science communication.
You can read more about the project, the team, and our publications on the official InSciM project page.