For most of the past year, Michela and I have been guest editing a special issue of the CERN IdeaSquare Journal of Experimental Innovation (CIJ) on the subject of co-creation. It’s finished now: twenty-six written contributions from around the world, plus our editorial, out shortly. And no, that is not a paid position, nor does it confer an honorary PhD in nuclear physics, but (in a phrase that will resonate with my musician friends) think of the exposure!
CIJ is a peer-reviewed journal of experimental innovation, published by IdeaSquare, the innovation space at CERN in Geneva. Michela was on their board of advisors for the last five years and we ran one of our labs there a few years back. CIJ is Diamond Open Access, which means everything in it is free to read, nobody pays to publish, and authors keep their own copyright. It also means I can share our editorial with you in full below, which is the point of this post.
The call for papers went out last September, and by the time submissions closed in March, there were far more than we could reasonably have expected. We imagined we’d get five. Ten at most. What followed was a year of the actual machinery of editing. There were four of us on the guest editorial team (Anne Dvinge and Ana Kuštrak Korper, alongside Michela and me), and we recruited nearly 50 experts (after approaching literally hundreds) so that each research article could be independently assessed by two external reviewers. Then the decisions, revisions, second rounds for the papers that needed them, and production that took us right up through August.
It’s been a huge part of what we’ve been up to for the last year, and it’s been escalating over the past few months, right up to the deadline. And we just brought it over the line, so I thought I’d share and then have a glass of something nice to celebrate. Enjoy.
Editorial: On co-creation
Andrew Dubber and Michela Magas
Guest editors, CIJ Special Issue on Co-creation
In a famous thought experiment, the philosophers Andy Clark and David Chalmers (1998) imagined Otto, a man with failing memory who carries a notebook holding what he can no longer remember. Because he consults the notebook as reliably as the rest of us consult our own recall, they argued, the notebook is not an aid to his mind but part of it. The argument could therefore be extended that you do not “collaborate” with an external intelligence source that you routinely deploy any more than Otto “collaborates” with his notebook. Intelligence has never been confined to the inside of the head. It has lived in notebooks, libraries, instruments and other people for as long as those have existed, and the extended mind now has a strange new annexe. From that perspective, working with AI is, in an important respect, not “co-creation” at all. It is simply creation.
McLuhan (1964) asserts that the content of any new medium is its predecessor. In the case of LLM model training, this is quite literally true, and not just a useful metaphor for unfashionable media studies theorists. There is a collaborative relationship at play, but it operates at a deeper layer, with the vast body of human work swallowed whole by the training data. You are not collaborating with AI. You are collaborating with everyone whose writing, speech and imagery the model has absorbed. Most of them were never asked, and none of them is credited, but you are collaborating with them all the same.
Yet the papers in this issue push back on both positions, and they are right to. In the pages that follow, a glacier holds a conversation. Harry Yeff’s paper describes an agentic voice system: an AI that speaks and listens, grounded in the research literature on the Thwaites Glacier in West Antarctica. At the AI for Good Global Summit in Geneva, it held more than 3,000 conversations, adapting what it knew of ice loss and sea level to each listener in turn. Co-creation is plainly not restricted to transactions between humans. There is co-creation with other species and living systems, as the immunologists and the ocean-dataset builders in these pages suggest. There is co-creation with place, whether a garden at CERN or the waterscapes of Kumasi. There may be co-creation with intelligences that are built rather than evolved, which does not make them unnatural: tool-making is among the most natural behaviours of our species.
Which leaves a genuine puzzle. How can we be in collaboration with an intelligence that exists beyond us but is also, on the prosthetic view, as much a part of us as Otto’s book is of him? Perhaps the simplest answer is that there is no conflict to resolve. The study of elusive, dual-natured phenomena is CERN’s trade. Physics made its peace long ago with entities that are two things at once, particle and wave, depending on how you look. The intelligence we now work with is at once our own, an extension of the collective human record, and separate from us, a sounding board with our own voice in it (the “pet self” as Yeff described it in our MTF Labs in Stockholm back in 2018).
This is why three pieces in the Journal’s conversational Coffee Paper format challenge us to consider how else we can co-create both with other humans and beyond. John Wood’s glossary of co-creation runs from abductive reasoning to his own coinage entredonneurship, ‘giving from between’ rather than ‘taking from between’, and pauses to calculate, via Euler, that a quartet offers six times the potential synergies of a duet. Marta de Menezes and Luis Graça, artist and immunologist, argue that art and the immune system are structural cousins: adaptive systems that turn encounter into development, draw robustness from diversity rather than uniformity, and are both, in their lovely phrase, poised toward what has not yet arrived. Kat Austen begins on a bus in Seoul and ends on the Irish coast, where around 100 handmade images from Galway coastal communities became the training data for a generative model trained entirely on renewable energy. The model helps people envision a liveable ocean future rather than remix the world as it is. Her question deserves to be quoted whole: “What if we build AI not on the fragments of the shape of the world that is, but on the shapes of the worlds to which we collectively aspire?”
From conversations with glaciers to grafting one’s partner’s cells on one’s body: that is the span of this issue. The question is who gets to make knowledge, with whom, and on whose terms. The range of the issue starts with definitions, then people and power, then places and structures, then machines, and ends with hope. This issue is about making knowledge together, across boundaries that often keep the makers apart. It is, by quite some distance, the largest issue this Journal has yet published.
A cluster of papers takes co-creation to places where the stakes are immediate, and the power relations are not abstractions. Barcham’s study of extended reality prototypes for climate resilience is written from the inside, from the author’s own Māori hapū (clan) in Hawke’s Bay. It shows how even a well-intentioned participatory process can reproduce colonial dynamics when data, platforms and resources remain in external hands. Lauterbach and colleagues hand the research process itself to musicians: an established highlife band and six community musicians in Kumasi, Ghana, who conducted their own fieldwork and co-wrote a song about urban water, on the understanding that a lyric can carry knowledge about living with polluted water that no policy document can convey. Saienko and colleagues work with six third-sector organisations in Dundee for whom, as one participant put it, “Data work is not a part-time undertaking”, and use participatory design both to surface and begin to redress that asymmetry of data power. Costa and colleagues take us into the boxing gym, where an embedded researcher, a former competitive boxer and referee, mediates between the mouthguard data and the lived expertise of athletes whose sport’s health evidence was built around men. Bugler and colleagues, working from a co-design project with young adults in Cape Town, ask a question every participatory researcher will recognise: when the workshop is over, and the analysis begins, can anyone retrace how the participants’ words became the researcher’s findings? Their answer, a discipline of traceability on a shared digital whiteboard, is one of the transferable and applicable gifts of this issue.
A second cluster examines the rooms and structures in which co-creation is staged. Suominen and Jussila look back at 17 educational hackathons in Finland and find that, across their cases, the ideas that survived the weekend were those where an external problem owner was present and the event had a deliberate afterlife. Their evidence includes a student concept, brewed from a food company’s surplus cucumbers, that ended up on shop shelves, and a recurring failure they call the torso problem: a well-built event with no afterlife. Bonucci and colleagues compress co-creation into 90 minutes at the Brussels final conference of ATTRACT, the EU initiative to turn big-science technology into applications. Even under those constraints, they show that 25 mixed teams can produce comparable, analysable pathways from big-science resources to societal challenges. Hofmann and Alliss bring school students into MRI sequence development at a Fraunhofer institute in Bremen, with glass art as the bridge: the glass appears in the scanner as negative space, and the students’ reported understanding of the physics rose dramatically over two days. Beigel asks how the paradoxes of structured freedom shape co-creation in learning and teaching, where the invitation to co-create is itself a form of structure. Bartoloni and Strigari trace how astroparticle physics and space radiobiology built a shared practice out of what the literature calls boundary objects, artefacts that both sides can hold on to: a shared ontology, dose-effect models, curated datasets. The paper proposes tools for logging who decided what, and who is owed credit, in cross-domain work. Tanaka and Ito describe a project that lives permanently at CERN: a garden of suspended silk threads in lilac, wisteria and violet, designed with students from two universities in Japan through the lens of Japanese aesthetics and particle physics, begun in 2018 and completed through the pandemic, and the practical constraints of building anything at CERN. Clarke and colleagues propose that a participatory sound platform with some 10,000 users can be governed as an experimental data commons for health research, provided consent, provenance and attribution are engineered in from the start.
A large share of the submissions we received concerned artificial intelligence, and we accepted the strongest of them, in part, because they disagreed with one another. Gelner and colleagues pitted six engineers against an enterprise language model on a corpus of battery-test reports and found the machine to be faster, but more often wrong. Tellingly, it could not conclude that an answer simply was not in the documents, a judgement all six engineers made correctly. Silver, Morrow and De Paoli’s methodology draws a firm line: in the TALLMesh workflow for thematic analysis (the craft of finding the themes in a body of interviews and texts), the language model processes patterns but “does not function as an analyst or collaborator”, and accountability stays with the researcher. Ochiai proposes an AI convenor for multi-domain co-creation and, with the understanding that convening is a form of power, sets hard boundaries and escalation rules before facilitating anything. Zampirolo and colleagues treat the governance of AI itself as a co-creation problem, to be worked out by the people who share the dilemma rather than specified top-down. Shen and Iandoli teach students to prompt critically and watch their creative agency grow rather than shrink. Quiñones Gómez surveys professional designers and maps the roles and responsibilities they actually grant AI in their work, and finds that trust follows transparency rather than sophistication. Etezadi and colleagues’ experiment uncovers a creativity discount: teammates known to use AI are rated less creative, and disclosure makes it worse. While a small study, it should give pause to anyone drafting a disclosure policy. Riznyk proposes community-owned linguistic datasets governed by collective vote, so that the languages that feed the models remain the possession of their speakers. Finally, Yeff, whose glacier opened this editorial, makes the boldest claim of all: that adaptive, listener-calibrated conversation is itself a co-creative act, and that meaning made this way belongs to neither party alone.
=======
It is hard to imagine a better home for an issue on co-creation than this journal. CERN is itself among the great standing demonstrations of co-creation: between nation-states, between disciplines – art and science meeting at the highest levels – in pursuit of the smallest constituents of the natural world and their own powerful interactions. Its commitment to Open Science extends the collaboration across time and distance to people who will never visit Geneva. That practice runs through this issue too: the aggregated data summaries and governance protocols behind Clarke and colleagues’ participatory sound platform are already openly deposited in the Journal’s community on Zenodo, CERN’s open repository for research data, where anyone may build on it.
If the intelligence we now work with is indeed dual-natured, then CERN is the right context in which to examine it, and if anyone were to build a particle accelerator and collider for intelligence, IdeaSquare – CERN’s experimental innovation space from which this journal takes its name – is exactly where you might expect to find it. One of us served a five-year term on the Advisory Board of IdeaSquare, and recognised in it something we knew from first-hand experience: what it takes to give a co-creation community a home.
We have over sixty years of co-creation practice between us, across design, media, data-driven technologies and the industries built on them. Since 2012, first as Music Tech Fest and then as MTF Labs, we have run 26 international events where scientists, artists, engineers, inventors, industry experts, and policymakers spend an intensive week building things together. In 2023, we brought the lab format to IdeaSquare itself, gathering New European Bauhaus thought leaders in think-and-do sessions in person for the first time. The community that has grown around those nomadic events now numbers 9,000 people, and one of its defining features is that nobody in the room is an expert on everything; everybody is an expert in a particular knowledge domain, and we progress by connecting the dots under the banner of grand missions such as Ecosystem Living or JUST Data. The community of the MTF Labs ‘intelligence collider’ vastly exceeds the attendance at any one of its collisions, and if, as Wood has it, four nodes offer far more possibilities than two, then 9,000 is a much bigger number than it appears.
The MTF Labs community has yielded a non-profit research institute, the Industry Commons Foundation. This is where the strange new annexe of the extended mind gets its architecture. Our project GRAPHIA is building an open knowledge graph for the social sciences and humanities: a connected map of the field’s data that machines and people can both navigate. LUMEN is building federated discovery across scientific domains, so that a researcher in one field can find and use what another already knows. A language model on its own is a rear-view mirror, fluent about where we have been, but a poor navigational tool. Grounding it in verified knowledge, in facts and ontologies, is what makes it a trustworthy companion for where we are going, and that grounding is precisely what these infrastructures exist to build. In both projects, co-creation is a major research stream: it powers the innovation labs and is proving remarkable both as a way of generating new knowledge and as social glue for researchers working in diverse domains. We read Wnuk and colleagues’ account of the OPERAS Innovation Lab, five years of work drawing on 655 participants from 33 countries, with more than usual interest: OPERAS is a partner in both projects, and their co-creation workshop, in October 2025, ran at our Innovation Prototyping Lab in Zagreb.
Co-creation has become a working method within European research infrastructures, and it now turns up where nobody would once have imagined it. This includes large industrial AI programmes such as our project EuroFMX, where Test and Experimentation Facilities stimulate innovation co-creation processes, LLMs reference novel industrial ontologies, and Living Labs are deployed as a validation methodology. The value that emerges from all these experiments belongs to no single field and usually to no single person. That is the kind of co-creation we know first-hand, and it has informed our approach to this journal: a working familiarity with what happens in the space between knowledge domains.
It is also how this editorial team came together: we know our fellow guest editors, Anne Dvinge and Ana Kuštrak-Korper, through the MTF Labs community. Anne is Senior Innovation Advisor for the social sciences, humanities and the arts at UCPH Lighthouse, the University of Copenhagen’s innovation unit. She spent a decade as a music researcher, then founded Low-Fi, a platform that turned private living rooms into concert venues, with each event co-created by musicians, hosts, and audiences in cities from Copenhagen to Santiago. These days, her role is in helping researchers across the humanities translate their knowledge into impact. Ana wrote her doctorate on the role of service design in service innovation, a field in which co-creation is not a topic but the working method, and she now puts that research into practice as innovation coordinator and civic designer in the Mayor’s Office of the City of Zagreb, while keeping a research foothold at Linköping. Both know the lab process from the inside, both brought their own disciplinary depth to the editorial table, and each handled a substantial section of this issue from first assignment to final recommendation. We are extremely grateful to them both.
We make no secret that some of the authors, too, are known to us through that same community, and that over the years several have become friends. Every research article here nonetheless passed rigorous peer review, exactly as it would in any issue of any other serious journal, and the price is that we have had to reject work by thinkers we know, respect, and will likely work with again. The reviews were anonymous, the standards were the Journal’s, and friendship bought nobody an easier ride. The same discipline ran the other way: Matteo Vignoli, the Journal’s Editor-in-Chief, is a co-author on one of the papers in this issue, and took no part in any decision concerning it. We are grateful to Matteo for his guidance and to the wonderful Valeria Brancolini at CIJ, who kept things on track and kept us moving, while handling record numbers of papers for typesetting.
Our thanks go to Alexandra Antonopoulou, Tarsh Bates, Fergus Bisset, Johan Blomkvist, Martina Čaić, Wanjun Chu, Andy Dearden, Alessandra Di Pisa, Sara Donetto, Ivo Emanuilov, Jeanette Falk, Lene Asp Frederiksen, Harald Fredheim, Visda Goudarzi, Hanna Louise Grønneberg, Anne-Sophie Guernon, Stefan Holmlid, Mark-David Hosale, Jamer Hunt, Liesbeth Huybrechts, Petter Karlström, Iva Kostešić, Ian Lambert, Laura Aakjær Larsen, Tau Ulv Lenskjold, Sara Lenzi, Marianna Lubanski, Agathe Malbet, Peter Meister-Broekema, Samuel Moore, Andrijana Mušura, Alexander Nolte, Divya Persaud, Sebastian Porsdam Mann, Emily Rempel, Matteo Romanello, Gergana Romanova, Vitaliy Roud, Clara Saglietti, Ann Shivers-McNair, Lina Svensberg, Francisco Teixeira, Tine Thorup, Horácio Tomé-Marques, Claudia Westermann, Renee Wever, Åsa Wikberg-Nilsson and Anna Xambó Sedó.
Between them, they carried out the peer review for this issue, with several taking on more than one manuscript and many returning for a second round. Their time and care were given to work that was not their own. In an issue about co-creation, that fact is worth highlighting. Some of the work in these pages was carried through genuine difficulty, including bereavement, and we are conscious of what it took to bring it to completion. The special issue was co-created by the guest editors, the authors, the reviewers, the CIJ editorial team and, in a real sense, the subject matter, and that collaboration lives in the gaps between the articles as much as in any single text.
This collection, formed of diverse voices, experimental practices, researchers from outside the traditional academy, and methods built on inclusion, is deliberately positioned as a rebuke against the forces that simplify, brutalise and exclude, and in outspoken favour of the things they attempt to degrade: scientific understanding, intellectual pursuit, truth, beauty.
Acknowledgements
The guest editors’ work on this issue was supported by the European Union’s Horizon Europe research and innovation programme through the projects GRAPHIA (Grant Agreement No 101188018) and LUMEN (Grant Agreement No 101187940).
References
Clark, A., & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19.
McLuhan, M. (1964). Understanding media: The extensions of man. McGraw-Hill.
