Open Source: The Key to a Transparent and Collaborative Scientific Publishing Landscape
The scientific publication process often suffers from inefficiency and delays. Valuable knowledge that could save lives can take up to two years to be published and shared among scientists. This is primarily due to the lengthy peer review process, which involves a paper being reviewed by two to three scientists before acceptance. Let’s examine the current peer review process in more detail and explore opportunities for improvement.
Typically, authors must wait several months after submitting their paper for qualified reviewers to assess it. Reviewers are provided with a plain text form for feedback, lacking modern tools like inline commenting, bullet points, or images. As a result, reviewers spend excessive time pinpointing specific sections of the paper, resulting in lengthy and difficult-to-digest feedback.
The author receives these extensive reviews and must transform them into a to-do list for making the necessary changes. This process involves breaking down the reviewers’ feedback and assigning tasks to co-authors. The revised paper, along with a detailed response to each suggested change, is then sent back to the reviewers and the editor. This back-and-forth resembles a cumbersome mail exchange reminiscent of the 18th century.
These procedural inefficiencies and delays significantly contribute to the overall time required for publication. Scientists, already burdened with numerous projects, struggle to handle the cognitive load caused by extensive reviewer feedback. The need for improvement is urgent, as every moment lost in this process can potentially cost lives. Furthermore, accepted submissions often end up behind paywalls that restrict access for those outside of wealthy institutions.
Efforts to address these issues, such as the one led by the Curvenote team, are limited. The Curvenote platform was initially designed to provide a better reading and learning experience by enabling the creation of interactive manuscripts, as opposed to plain PDFs. The team is now working on incorporating peer-review capabilities into their platform, including improvements like organizing feedback as a to-do list and customizable ordering based on difficulty levels.
However, addressing only a part of the problem is insufficient. A completely new system based on practices from the open-source software development community is necessary.
Decentralized publishing
In this new system, manuscripts would be hosted in repositories with version control and collaboration features similar to those on GitHub. Researchers would decide whether to make their repositories public from the start or only publish them when the manuscript is ready for the research community. Tools such as Myst Markdown can generate interactive web versions of the manuscripts for easier reading.
Instead of being trapped behind profit-driven journal paywalls, these repositories could be hosted by researchers’ institutions. This approach, which we could call DeSci Publishing, would remove traditional journals from the equation.
Now that journals are out of the way, you probably think there are three considerable issues with DeSci Publishing:
How would papers get reviewed?
How would impact factor get measured?
How could interesting publications be found?
How would papers get reviewed?
When it comes to reviewing the papers, open-source practices could be applied. In open-source development, when users want to request a new feature for software or when they want to report a bug, they open an issue in the software’s repository. Well, if you think about it, peer reviewing is nothing more than feature requests and bug reports.
For instance:
“The manuscript should add an extra statistical analysis...” → Feature request
“The manuscript is missing the following citation...” → Bug report
“Figure 2 should show more examples...” → Feature request
“The statement made in line 87 is incorrect as shown by...” → Bug report
“I could not replicate the experiment presented in Figure 4...” → Bug report
Anyone would certainly have access to open, see, and interact with issues to critique and collaboratively improve scientific work. A user interface similar to that of GitHub’s pull requests, where inline comments and suggestions are allowed would be necessary. The writers would then open a pull request for those issues they want to work on and include in the manuscript, generating a new version of the manuscript when the pull requests are merged into the manuscript.
How would impact factor get measured?
The impact factor is an important metric for researchers, as it is often used by funding bodies to evaluate the scientific output of a researcher, a research lab, or a research center. Making sure that DeSci Publishing offers this metric is fundamental for this new system to work.
First, the impact factor could be measured similarly to how it is currently done by simply counting the number of references a manuscript has. Second, just like GitHub repositories can be “starred,” manuscript repositories could also be liked or starred. To avoid the use of bots to boost impact factor stats, the platform could only let users log in with their ORCID number.
How could interesting publications be found?
We live in times when every social media has its algorithm to customize its feed to our interests. Coming up with a platform that shows new articles, issues, or discussions on a feed as they happen should not be difficult. Researchers would be able to see items related to specific keywords they’re interested in, similar to how Bluesky feeds work.
For those who don’t like feeds, the platform could also provide an AI-powered search bar that would return papers answering specific questions, similar to what Elicit does.
Additional Benefits
Another important benefit would be the ability to cite specific lines or paragraphs of an article rather than citing entire manuscripts. Since the system would be built on a platform where links can be made to specific lines, it would be possible to cite specific claims so that a single click would take you to the original claim.
The use of repositories for manuscripts would also make sharing code very easy, making science more open and reproducible.
Challenges
Any new system requires training its users, and the learning curve could be steep for those unfamiliar with GitHub and open-source practices. The system would need to simplify the user interface and borrow UI patterns from social media. Another challenge would be educating researchers to follow these new practices.
Conclusion
I believe the different parts of the DeSci Publishing system I envision already exist independently in different settings like social media or software development. It is a combination of them upon a decentralized architecture that could make DeSci Publishing a reality.
The people who are crazy enough to think they can change the world, are the ones who do.
Steve Jobs