Reimagining scientific publishing
Science is at the core of our company, and we believe that open science is better science. While the traditional thinking is that profitability and openness are at odds, we think that sharing our research is, in fact, directly aligned with our goals. Science is an inherently collaborative craft — any scientist knows the benefits and impact of discussing their work with colleagues. Insights from those with differing expertise is often key to unlocking tricky problems and truly innovating. It’s also a great way to figure out whether your work is useful to others.
In light of this, one of the first decisions we made as a company was that we’d share a lot of our science, but not in journals or on preprint servers. We thought the best way to develop a publishing strategy fit for our goals would be to experiment, and we wanted to push beyond existing, narrow solutions. Though preprints solve many of the problems in traditional publishing by making work available earlier and without a paywall, we want to push further and explore different formats, displays, and feedback mechanisms. We want to think clearly about our practical needs and avoid letting the means become the end. That said, we’re also making note of what hasn’t worked in the past to make sure other publishing efforts inform our strategy.
By fully separating from established approaches, we’ve stripped away many assumptions, forced ourselves to define what matters to us more clearly, and expanded the range of what’s possible.
Evolving a new system through experimentation
We’re optimizing our approach around our own goals, but we ultimately hope to make meaningful change for others as well. There’s a dearth of options out there for those with similar goals, especially scientists at for-profit companies, and we want to provide ideas for anyone else not served by the current publishing landscape. Arcadia, which is untethered from the academic system, is in a unique position to leverage the science in-house to demonstrate what’s possible. We have an unparalleled opportunity to reset, learn from innovations already happening in the open science community, and construct something radically new.
We’re conceptualizing our publishing efforts as an “experiment” that will involve ideation, testing, analysis, and adjustments based on initial results. By repeating the process, we can iteratively improve our model. Critically, we’re doing this iteration closely with our in-house scientists, striving to design a new system that achieves our high-level goals and meets the practical requirements of the researchers themselves. We want them to feel a strong sense of agency and have the ability to discuss their work on their own terms, avoiding new obstructions to sharing. We still have high expectations about the quality of their work, and they’re responsible for thinking hard about how publishing ladders up to our big-picture company goals. Our scientists will learn how to drive scientific dialogue on their own, a skill that will serve them throughout their careers. Further, we’re sharing our plans, progress, and results here so others can weigh in and learn from our work, just as we hope will happen for our scientific research.
Our central hypothesis
While we want to make an impact beyond Arcadia, we think timely, open sharing will help us meet our goals as a for-profit science company. We hypothesize that working openly and routinely engaging with the rest of the scientific community will help expand our thinking, influence our direction, and speed our progress. There will be some research we hold back when essential for commercialization, but the vast majority of our work (especially tools and early-stage discovery efforts) can be accelerated and more useful when shared and discussed publicly. If we’re able to validate this hypothesis, we’re hopeful that other companies will appreciate the value of the model and adopt it, improving science as a whole.
Goals
After assessing the key limitations of the current system and our goals as a company, we’ve pinpointed three key qualities to maximize in our publishing strategy:
- Speed. Sharing smaller, more modular pieces of research as we go will let people learn about and use our findings quicker and will speed scientific progress as a whole.
- Utility. By breaking from rigid journal formatting, we can maximize usability and explore interactivity. Our data will be easy to find, access, use, and repurpose in ways we can’t predict.
- Rigor. We need public comments from anyone. Expertise lives everywhere, not just where you look for it. With diverse feedback and iterative engagement, our work will be the best it can be and we can meet community needs.
Progress
Since we wanted to share ideas as quickly as possible and let others try our publishing approach, we decided to use a pre-existing platform to host our first research products. We chose the platform you’re on now — PubPub!
For an overview of why we chose this platform and how we set things up in this first iteration, check out our intro piece:
How can we measure the value of science?
Traditional metrics fail to reflect the true value of scientific work, especially in open science initiatives like ours. We’re seeking feedback on new ways to measure and communicate the impact of research! Share your ideas to better assess the quality and usefulness of openly published work, and inspire others to rethink how we do this for publications across the board.
Iteration 2.0: Scientists fully drive their pubs
Fifty pubs and two years into our publishing experiment, we’d made many improvements to our approach. But this inadvertently led us away from our original vision of scientists sharing their ideas, findings, and challenges at a pace that can accelerate their research.
In summer 2024, we committed to a new phase of our experiment in publishing — shifting from a more top-down approach to a bottom-up framework where our scientists are expected to decide how and when they share their work. This means scientists need to fully own the publishing process and take a more proactive role in thinking about their communication strategy as well as scientific and commercial implications around sharing. As long as they include the code, data, and other content needed to replicate the work and all byline contributors sign off, we’ll release it.
In the following pub, we share some observations and perspectives from our co-founders and publishing team. We also lay out the basics of what this next chapter of our open science experiment is testing:
Six months into model 2.0, almost everything is better, but pubs are slower
Now that we’ve tried this more scientist-led approach for a little while, we decided to take stock of how it’s been going. We’re hitting our major goals: we’ve increased scientist agency and experience with the process while reducing meetings and maintaining pub quality. Our scientists take their responsibility to make decisions on commercialization seriously and front-load those discussions such that they haven’t been slowing pubs down. With less time spent on pre-determined process steps that we’ve been able to cut, our pub team has more time to assess and improve the overall publishing model. The biggest issue is that pubs take longer without top-down oversight and project management, so we share less often. We also haven’t seen any improvement in outcomes from engaging with the outside world — our scientists seem to engage less often now, we see a bit less interaction with our work, and the usefulness of the feedback we get via public comments has stayed the same.
Some of these conclusions are based on limited data, so we hope to both update this pub and release others when we can say more. Overall, we’re pleased with what we see and plan to continue along the same track. We have plenty of ideas to overcome our challenges, but would also appreciate feedback from readers.
For all the details and more extensive reflection, check out our first v2 progress report and weigh in with your thoughts:
Trying a new format — computational notebook pubs
Most of our research so far has been computational, and much of that work happens in Jupyter Notebooks. Since this is where we do a lot of our analyses and often how we share that information internally, we realized we could publish faster if we could convert these notebooks directly into publications. This cuts out the extra steps of writing up a separate pub to accompany a GitHub repo, hopefully speeding the publishing process for our scientists. Moreover, since we’re sharing all the code that generates the pub, it makes the entire product fully reproducible so others can easily build on the analysis.
There’s been interest in so-called “executable papers” for years, and though not mainstream, they’ve been adopted to some degree. We were glad to see that we had multiple technical options for implementing our own version, and chose to build with Quarto. Ultimately, we’d love to be able to host notebook pubs right here on PubPub. We’ve shared our thoughts on the concept and a template you can use to try it yourself in the pub below:
What’s next?
In 2025, we’re excited to migrate to PubPub Platform and rethink how we present pubs and higher-level site content once we have more control. We'll experiment with turning Python notebooks into pubs and explore how AI can potentially increase the quality of our work. Longer-term, we’d like to enhance discoverability, facilitate other groups using the platform and sharing more openly, and develop new ways to measure our impact.
