“If you build a data set and nobody can find it, is it useful?” asks a new Nature news feature published July 1. “Not as much as it could be.” The question is rhetorical but the stakes are not: with trust in science under sustained pressure from partisan actors and viral misinformation, the accessibility and transparency of scientific information have become a matter of institutional urgency.
The article examines the state of the FAIR data principles a decade after they were first drafted. FAIR, Findable, Accessible, Interoperable, Reusable, is a framework designed to ensure that research data can be located, opened, combined with other datasets, and used by others long after the original study is published. The principles were formally published in Scientific Data in 2016 by Wilkinson and more than 40 co-authors under the FORCE11 umbrella, and the paper has since accumulated roughly 16,000 citations.
Barend Mons, the Dutch molecular biologist who conceived the FAIR framework, told Nature that its core purpose is accountability. “The more the data are understandable by people other than the creators, the more we are able to determine not only the trustworthiness of the data set itself, but also its alleged creators.”
What FAIR Actually Demands
An ideal FAIR dataset, as described in the article, is properly documented with detailed metadata, when, where, and how the data were created. It is designed before data collection begins, with data-management plans that specify appropriate licenses and persistent identifiers. It is structured so that both computers and humans can find it and integrate it with other datasets.
The principles are deliberately general. As Mons noted, FAIR “cannot address the specifics of every application.” Other researchers have since extended the framework to cover algorithms, research software, and open-source projects through initiatives such as FAIR4RS and FAIR-USE4OS.
FAIR in Practice
The article highlights several discipline-specific implementations. In high-energy physics, the FAIR4HEP project has evaluated Large Hadron Collider data for FAIR compliance. Carnegie Mellon University has released dedicated FAIR guides for chemistry, mathematics, neuroscience, and psychology. The British Ecological Society published a guide to reproducible code rooted in FAIR principles. In artificial intelligence, the HuggingFace platform promotes “model cards” that document intended use, performance metrics, training data, and limitations.
“In many ways, it’s like cooking,” said Amelia Jimenez-Sanchez, a data-integrity researcher at the University of Barcelona. “Once you have the right ingredients, or familiarize yourself with FAIR practices, it becomes easier to make a meal. Eventually, it just becomes a part of how you do your work.”
Natalie Cooper, a macroecologist at the Natural History Museum in London, emphasized that data alone is not enough. “Data are data, but there’s also the entire system of infrastructure that is built around it to store, share and analyze that information, and those tools need to be fair and reproducible too.”
Neil Chue Hong of the Software Sustainability Institute at the University of Edinburgh added: “It’s now very hard to analyze or visualize data without software, and at the same time, it’s very hard for software to exist without high-quality data.”
Policy Momentum
Governments, funding agencies, and publishers have increasingly made FAIR-compliant data sharing a requirement. The Australian Research Data Commons offers a FAIR Data Self-Assessment Tool that provides practical guidance for improving FAIRness. Discipline-specific FAIR resources now exist for astronomy, materials science, genetics, and single-cell genomics.
The article appears alongside a companion piece in Nature asking whether trust in science has genuinely declined, and the answer, based on survey data, is more complicated than simple narrative suggests. But the FAIR framework positions itself as a structural response: not a campaign to persuade skeptics, but a technical infrastructure that makes science verifiable by design.
At 10 years old, the concept has moved far beyond its origins in bioinformatics. Whether it can live up to its founders’ ambition of embedding reproducibility into the DNA of scientific practice will depend, in the end, on whether researchers across fields are willing to learn the recipe.