23 August 2017

Communicating uncertainty in research papers

Posted by Shane Hanlon

By Madeleine Jepsen. This is the first of two posts on communicating uncertainty.

Uncertainty is an important part of research, and scientists often carefully plan out which types of uncertainty analysis are suitable for their research. However, one important part of uncertainty — how it is verbally framed in the paper — is often overlooked in scientific disciplines. Joseph Guillaume, a postdoctoral fellow at Aalto University, said he was inspired to look into this issue while doing research on global water resources. While the research in this field was accepted and considered valid, he said traditional methods for uncertainty analysis were not being used, and indeed did not seem suitable. So how were the scientists dealing with the uncertainty present in their work?

“That was the core idea I wanted to look at, and I realized that it was in the publication itself that this would probably come out, because if you look in detail at what was written, you can pick up the hints that authors have given about how they think about uncertainty,” Guillaume said.

In the research, Guillaume and colleagues examine paper abstracts from Water Resources Research to analyze the different ways scientists communicate uncertainty surrounding their claims.

They came up with five core questions that deal with uncertainty in a paper:

  1. To what extent is the conclusion ready to be used?
  2. What limitations are there on how the conclusion can be used?
  3. How certain is the author that the conclusion is true?
  4. How thoroughly has the issue been examined?
  5. Is the conclusion consistent with the reader’s prior knowledge?

For any given claim in a paper, the answer to these questions lies in the wording of statements in the paper. The uncertainty therein can be portrayed as nonexistent (where the claim is expressed as fact), overwhelming (where the claim is presented as an opportunity to learn from the researcher’s failure), or anywhere in between these two extremes.

Though Guillaume’s research focuses only on analyzing papers’ abstracts, his recommendations are worth considering for authors and peer reviewers:

  • Consider the different ways uncertainty surrounding a claim can be communicated — there are many ways to accomplish this, and some work better than others depending on the situation.
  • When possible, look for generalized conclusions that can be made in order to prompt useful debate. The proper framing of uncertainty can help make more generalized claims possible. Sometimes, including speculation can be helpful for future projects if it’s properly framed.
  • Intermediate-level claims can prevent researchers from both over-inflating confidence in a claim or from rendering claims meaningless in throwaway lines with vague references to the need for future work.

Guillaume said there is typically a disconnect between sociological research on uncertainty framing and the application of these ideas when scientists write about their research for publication. His research on uncertainty is meant as a conversation-starter about an important part of science that is sometimes neglected.

“There’s this whole field of sociology of science, which looks at the question of ‘How is it that we go from all this social activity to something that is accepted as true?’” Guillaume said. “That’s obviously fascinating, but much like any science, what they’ve looked at doesn’t really filter down to me as an individual scientist unless I deliberately go looking for it.”

Madeleine Jepsen is a public information intern at AGU, and Joseph Guillaume is a postdoctoral fellow at Aalto University, Finland, and visiting fellow at the Australian National University. Joseph’s uncertainty research can be found here, and more information about uncertainty in scientific research can be found at The Society for Decision Making Under Deep Uncertainty’s website, and at the website for the AGU hydrologic uncertainty technical committee, responsible for coordinating and promoting  sessions on hydrologic uncertainty at AGU’s 2017 fall meeting.