Open and participatory approaches to science have been expanding in recent years and are increasingly required by scientific funding bodies. But do these practices only have positive impacts? Are there particular conditions or caveats that need to be considered?

These were some of the questions of the track Open, Collaborative and Participatory Science: Rethinking Knowledge Legitimacy, Policy and Science Futures at this year’s EuSpri conference, which brought together researchers from open science, citizen science and digital commons to share their experiences. I would like to thank Mayo Fuster Morell for making the track happen and for inviting me to discuss my thesis project in this context. In the following, I will summarize some of the discussions and ideas that emerged over the three days of the conference.

Open and participatory science

To begin with, a quick refresher on what I mean by open and participatory science.

Open science is an approach to research that aims to share “knowledge, results and tools as early and widely as possible”. This means, first of all, publishing research results, particularly articles, in open access journals so that anyone can read them free of charge. It also includes practices such as sharing research data and code, preregistering hypotheses and methods, open peer review and publishing preprints. The aim is both to make scientific knowledge accessible to more people and to improve the transparency, quality and reproducibility of research. Funding bodies such as the European Commission and France’s National Research Agency increasingly require or encourage open science practices as a condition of funding.

Citizen science or participatory science, by contrast, opens up the research process itself to people who are not professional researchers, allowing them to participate actively in scientific knowledge production. This can take many forms and involve different levels of participation. People might, for example, label datasets for research teams, performing tasks that computers are not yet good enough at, such as transcribing handwritten texts or identifying animal species in wildlife footage. They might also collect wildlife sightings in their local environment for species conservation, or contribute their lived experience to research on health conditions. The motivations are equally diverse. Researchers may use citizen science to mobilise large numbers of people to expand their capacity to collect or process data. Communities themselves may initiate projects to generate evidence about issues affecting them and use it to advocate for their rights. While citizen science is less commonly a funding requirement than open science, funding bodies increasingly encourage or reward it through dedicated programmes or evaluation criteria, seeing it as a way to make research more responsive to societal needs and strengthen relationships between science and society.

How generative AI challenges open and participatory science

Opening access to resources and processes affects, and is affected by, the environment in which they exist, creating new questions and challenges. The track discussed several examples of tensions that have recently emerged, in particular with the advent of generative AI.

The first concerns the massive scraping of open-access articles for LLM training and by AI agents, exemplified by arXiv, in a talk by Jean Constantin. ArXiv is an open-access preprint server where researchers can publish their work before peer review. With millions of academic papers, it is a valuable resource for training AI systems. While arXiv generally permits computational use of its content, the scale of scraping puts significant strain on its limited infrastructure, creating financial and practical burdens for the nonprofit organisation.

Generative AI does not only affect how content on arXiv is used, but also what is contributed to it. Since the beginning of 2025, arXiv has seen a dramatic increase in article submissions, mainly due to “AI slop” contributions, particularly in the computer science category. ArXiv relies on volunteer moderators to ensure that submissions are appropriate and on topic, and the influx has made this increasingly difficult. As a result, arXiv has introduced new requirements, namely peer review for certain categories of papers, as well as endorsement by an established author for first-time submitters.

AI is also changing how people participate in citizen science. In Zooniverse, the largest citizen science platform, research teams provide datasets that volunteers annotate. The astronomy project Galaxy Zoo has introduced a bot that automatically labels easier galaxies and only transfers the trickier ones to human annotators. This can increase productivity and remove some tedious work from human participants. However, it also risks taking away opportunities for newcomers and users with less expertise to learn by working through easier cases, making the project more geared towards expert users. See this editorial co-authored by track participant Marisa Ponti for an overview of use cases and ethical considerations of AI in citizen science.

Another source of tension can arise between project staff and contributor communities, as illustrated by iNaturalist. iNaturalist is a citizen science platform and social network for sharing wildlife observations, with more than 350 million observations as of August 2026. This year, the organisation joined a Google grant programme to integrate AI into its tools. The decision triggered a backlash among some volunteer contributors, who objected to the change for reasons including, for example, the environmental impact. A central issue was not only the introduction of AI itself, but also how the decision was made: contributors were not consulted and had no formal way to influence it, leaving them with the option of leaving the platform and deleting their contributions.

These examples illustrate several challenges that arise from the confrontation between open and participatory practices and the growing use of generative AI: increasing infrastructure and moderation burdens for volunteer and nonprofit structures, largely caused by for-profit companies that profit from these resources without adequately giving back; tensions between AI adoption and community values, highlighting a lack of mechanisms for integrating communities into decision-making; as well as trade-offs between productivity through automation and other objectives of citizen science projects, such as learning and providing enjoyable tasks for participants.

Openness alone is not sufficient – lessons from the history of digital commons

What these examples illustrate is that simply opening up resources and processes is not enough. A healthy knowledge system also depends on how these resources are created, maintained and governed. This is where the concept of the commons can provide a useful perspective.

In her seminal work Governing the Commons (1990), Nobel prize winner Elinor Ostrom distinguishes between the nature of a resource and the collective action through which it is managed. She distinguishes open-access resources, to which no one has defined rights or responsibilities, from common-pool resources that are collectively managed by a defined group according to shared rules for access, use and maintenance. Without mechanisms for collective management, common resources risk being depleted or degraded. This shifts the focus away from the resource itself and towards the activities that sustain it: creating, maintaining, governing and protecting it. In other words, it shifts the focus from the commons as a resource to “commoning” as an ongoing practice.

This year also marks the 25th anniversary of the term “commons-based peer production”, first used by Yochai Benkler in 2001, and the 20th anniversary of his landmark publication The Wealth of Networks. A quarter of a century ago, access to information technologies was expanding rapidly, allowing individuals around the world to access and contribute to knowledge creation at relatively low cost. This contributed to the emergence of new forms of distributed production, most notably open-source software and Wikipedia, created and maintained by communities of individuals outside traditional commercial structures. Based on these developments, Benkler envisioned the expansion of commons-based peer production as a potential basis for a radically decentralised, non-market-driven model of information production, one that could contribute to more egalitarian societies and a redistribution of wealth and power.

Peer production did become hugely successful, but not quite as envisioned. Alongside community-owned and governed projects, it also contributed to the emergence of the platform economy: a highly centralised internet dominated by a relatively small number of large commercial platforms. These platforms rely heavily on content and participation generated by their users, while retaining control over the infrastructure and capturing much of the value created through it. Coordination and visibility are increasingly determined by opaque algorithms, while user interactions are monetized for the benefit of the platforms. The arrival of generative AI can be seen as a culmination of this dynamic, with vast amounts of openly accessible knowledge being extracted from the internet commons and enclosed within commercial products.

The parallel was raised during the track as a point of reflection: many of the questions now emerging around open science echo debates about the information economy from two decades ago. Can the history of digital commons teach us how to preserve the benefits of openness and participation without allowing scientific knowledge to become primarily a resource for extraction?

Ways forward

So what are the ways forward? What might a healthy science commons require? I would like to point to three areas.

Who gets to decide?

The first is collective governance. If communities do most of the work of creating and maintaining a commons, they need to have a meaningful say in how it is governed, rather than being limited to “voting with their feet” and leaving.

Polski and Ostrom (1999) provide a useful framework here. Beyond the operational level, meaning the day-to-day actions people take when contributing to a commons, such as adding a wildlife observation to a database, there is the collective-choice level, which concerns how operational rules can be changed, and the constitutional level, which concerns how these rules can themselves be changed. Making these levels of decision-making explicit early on can help prevent conflicts between communities and those managing their commons. See the track contribution of Bastian Greshake Tzovaras for further reading on how different digital commons have addressed this issue.

Who contributes, and who benefits?

The second question concerns the imbalance between contribution and extraction, which remains a major challenge. One approach is to establish conditions for reuse through licensing, although licenses may be ignored, as has happened with content from Wikipedia used by big tech companies ignoring copyleft license conditions for derivative works. Another is bounded or moderated access, making resources available only to people or systems that agree to certain conditions of use. This can protect a commons from unwanted forms of extraction, but also means giving up some of its openness. Communities can also define conditions for contributing to the commons, for example by regulating or prohibiting the use of AI tools in contributions. This is a path that a growing number of open-source communities are exploring, for example Codeberg.

Who maintains the infrastructure?

The third question is who maintains the infrastructure. Knowledge commons may be “immaterial”, but they depend on very material infrastructure: servers, platforms, technical maintenance and long-term funding. During the track, it was suggested that universities could take a stronger role in providing this support for scientific commons, ensuring stability so that knowledge commons do not disappear when short-term project funding runs out. The scientific archive Zenodo, for example, is operated by CERN, proclaiming “Your research is stored safely for the future in CERN’s Data Centre for as long as CERN exists”.

Conclusion

Open and participatory science are valuable approaches to creating and disseminating scientific knowledge. But simply opening up scientific processes and results is not sufficient: openness is only one dimension of a healthy knowledge system. Commons theory offers a useful perspective by shifting attention not only to the “commons” as a resource, but also to “commoning” as the practice of creating, maintaining and governing it. This also means thinking about how communities can protect the resources they maintain from forms of use that undermine their interests, and how they themselves can participate in decision-making. Additional questions that need to be addressed when opening up resources and processes are therefore: who gets to decide, who contributes, who benefits, and who maintains the infrastructure?

References

Haklay, M., Fraisl, D., Tzovaras, B. G., Hecker, S., Gold, M., Hager, G., … & Vohland, K. (2021). Contours of citizen science: a vignette study. Royal Society open science, 8(8), 202108.https://doi.org/10.1098/rsos.202108

Greshake Tzovaras, B. (2025, June 12). Generative “AI” in citizen science: The iNaturalist backlash. https://doi.org/10.59350/7ykzb-dkx88

Fortson, L., Crowston, K., Kloetzer, L., Ponti, M. (2024). “Artificial Intelligence and the Future of Citizen Science,” Citizen Science Theory and Practice 9(1), 32. https://doi.org/10.5334/cstp.812

Ostrom, E. (2015). Governing the Commons: The Evolution of Institutions for Collective Action (1st ed.). Cambridge University Press. https://doi.org/10.1017/CBO9781316423936

Benkler, Y. (2006). The wealth of networks: How social production transforms markets and freedom. Yale University Press.

Polski, M. M., & Ostrom, E. (1999). An institutional framework for policy analysis and design. In Elinor Ostrom and the Bloomington School of political economy: A framework for policy analysis (Vol. 3, pp. 13–48).

Greshake Tzovaras, B. (2026, February 10). Open science and commoning beyond licensing: Deteriorating digital commons in the absence of collective governance. https://doi.org/10.59350/14dpd-23089