If you create a data set and no one can find it, is it useful? Not as much as it could be. With trust in science under siege from partisan actors and impartial pathogens, the accessibility and transparency of (and trust in) scientific information must be improved.

Have people stopped trusting science? The data tells a surprising story

Introduce FAIR data principles. In 2014, scientists realized that data management and stewardship could benefit from a set of shared guidelines, and dozens of international researchers came together to draft new recommendations. The resulting principles, which stated that data should be Findable, Accessible, Interoperable and Reusable (FAIR), were published ten years ago.¹. The original publication has around 16,000 citations, and governments, funders and publishers around the world are now calling for the data to be hosted and shared in a FAIR-compliant manner.

However, a decade later, even the founders recognize that the FAIR principles are an imperfect tool. Barend Mons, a molecular biologist at Leiden University in the Netherlands who conceived the initiative, says FAIR was always intended to be a set of general principles, “and therefore, by definition, cannot address the specifics of each application.” Fortunately, other researchers have taken the framework and expanded it to cover the broader data ecosystem.²including the algorithms, tools and workflows that drive contemporary research.

Make each discipline FAIR

At its core, FAIR aims to ensure that data is produced, analyzed, stored and shared in ways that promote transparency and reproducibility. “The more people other than its creators understand the data, the more we can determine not only the reliability of the data set itself, but also its supposed creators,” says Mons.

The ideal data set should be properly documented and easy for both computers and people to find and use. It should also be easy to integrate with other data. To achieve this, scientists must design workflows before data is collected and create and maintain a detailed metadata file, an often overlooked component that contains contextual information about the data set, such as where and when it was created. The initiative also prioritizes data management plans, including choosing appropriate licenses and persistent identifiers (the unique tags assigned to different resources) so that any information created during a project can be found and used long after the investigation is completed.

The complex truth about trust in science

“There’s a lot to think about, and I can see why it might seem really daunting for some scientists to consider it,” says Amelia Jiménez-Sánchez, a data integrity researcher at the University of Barcelona in Spain. But FAIR is like cooking, he says: Once you have the right ingredients, or become familiar with FAIR practices, it becomes easier to prepare a meal. “Over time, it becomes part of how you do your job.”

Users have adapted these practices to their disciplines. Carnegie Mellon University in Pittsburgh, Pennsylvania, has published FAIR guides for chemistry, mathematics, neuroscience, and psychology. Other initiatives have focused on astronomy, materials science, genetics, and single-cell genomic data. For fields without dedicated FAIR resources, researchers in the Netherlands have published “ten simple rules” to start conversations about FAIR practices.³.

Recognizing that discipline-specific resources did not exist in his field, Eliu Huerta, a theoretical physicist at Argonne National Laboratory in Lemont, Illinois, began adapting FAIR principles to high-energy physics. Today, Huerta is part of a collaboration called FAIR4HEP, which aims to help researchers in the field improve their data sharing practices. In 2022, he co-wrote a study evaluating data from the Large Hadron Collider at CERN, the European particle physics laboratory near Geneva, Switzerland, for its “FAIRness.”⁴. Among other things, the study “provides a set of step-by-step, domain-independent checks to guide the process of making a data set FAIR,” it says, a process the authors call FAIRification. The web-based FAIR data self-assessment tool from Australian Research Data Commons, a company building research data infrastructure in Melbourne, also offers “practical advice on how to improve the FAIRness” of your data.

Expanding beyond data

The FAIR guidelines also apply to software. The FAIR-USE4OS guidelines⁵ extend FAIR principles to open source software projects, for example, and initiatives like FAIR4RS focus on research software⁶.

“Data is data, but there is also the whole infrastructure system that is built around it to store, share and analyze that information, and those tools also need to be fair and reproducible,” says Natalie Cooper, a macroecologist at the Natural History Museum in London.

Six ways to put the public at the center of science and policy

Last year Cooper edited a guide to reproducible codes on behalf of the British Ecological Society which is based on FAIR principles. The code and data share many characteristics, so many of the recommendations remain the same. But something he’s found most useful in his own work is code review, which Cooper now does before submitting anything for publication. During the review, colleagues exchange protocols, test their reproducibility, and suggest ways to improve efficiency. “You just give each other feedback and hopefully you can improve each other’s code,” Cooper says. “It can be a really positive experience.”

Neil Chue Hong, founding director of the Software Sustainability Institute at the University of Edinburgh, UK, helped establish the principles of FAIR4RS. Hong says that in recent decades, the increased reliance on software is one of the biggest changes in data science, so that almost all branches of research now use software in some way. Therefore, the institute defends the fundamental importance of training scientists on best practices in the use of research software. “It is now very difficult to analyze or visualize data without software, and at the same time, it is very difficult for software to exist without high-quality data,” he says.

Just as data should come with a metadata or README file that contains information about the data set itself, software and algorithms should also have good documentation, including the version a person used. This is especially true for artificial intelligence research. For example, HuggingFace, a New York City-based model sharing service, encourages researchers to create “model cards” that provide key information about AI tools, including their intended use, performance metrics, training data, and limitations.