TL;DR
This paper argues that the current scientific practice of publishing research often fails to preserve the underlying data and associated workflows, hindering reproducibility and future discovery. It proposes ten actionable rules for scientists to better manage, share, and cite their data, emphasizing the long-term value of well-cared-for scientific information.
What Problem Does This Paper Address?
In today's research landscape, the publication of a journal article is often considered the end of a project. However, unlike historical scientific records like Galileo's detailed astronomical observations, modern publications rarely include enough information to fully reproduce or even verify the results. This lack of accessible data and detailed workflows makes it incredibly difficult for other researchers to build upon existing work, leading to lost scientific progress and wasted effort. Furthermore, scientists often struggle to even retrieve their own data years after publication. The paper highlights the growing complexity of "data" itself, encompassing not just observational or experimental results but also outputs from simulations and massive community-driven databases, blurring the lines between data, code, and analysis.
Methods
This paper is not a traditional research paper presenting novel algorithms or experimental results. Instead, it offers a set of guiding principles and practical advice for scientists. The "methods" employed are essentially best practices and recommendations for data management, sharing, and citation. The authors draw upon historical examples (like Galileo) and current challenges in scientific data handling to formulate these rules. They advocate for a paradigm shift in scientific culture towards greater openness and reusability of research outputs.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Data is more than just numbers: The paper emphasizes that "data" now encompasses simulations, databases, and even the code used to generate or analyze them.
- Galileo as a model: Galileo's meticulous record-keeping, integrating data, metadata, and text, serves as an early example of good scientific data practice.
- The "repeat-play" nature of data management: Caring for your data now benefits you later and encourages others to do the same, fostering a more collaborative scientific ecosystem.
- "Permanent" identifiers are crucial: While nothing is truly permanent, persistent identifiers like DOIs are essential for long-term data access and citation, moving beyond fragile URLs.
- Provenance is key to reuse: Understanding how data was generated, processed, and analyzed (the "workflow") is as important as the data itself for reproducibility and reuse.
- Sharing is not just about altruism, but also credit: The paper addresses how to ensure proper attribution for data and code, moving beyond traditional citation practices.
- The culture needs to change: Beyond technical solutions, a fundamental shift in scientific attitudes is needed to value data sharing and management as highly as publications.
Key Results
Not applicable, as this paper is a set of recommendations rather than an empirical study. The "results" are the proposed ten rules themselves, designed to improve scientific practice.
Limitations and Future Work
The authors acknowledge that implementing these rules requires a cultural shift and may involve extra effort. They note that while institutional approaches are important (e.g., data repositories), changing individual scientist's mindsets is also crucial. Future work is implicitly suggested by the ongoing development of standards and tools for data citation, attribution, and repository management, as highlighted in the "Links to Useful Resources" section. The paper encourages scientists to advocate for better data practices within their disciplines and institutions.
Summary updated