Section 1 | Introduction: The data sharing institutionalisation challenge

In Europe today, data sharing is more than a technical issue and has emerged as a site of political experimentation and regulatory ambition (Coche et al., 2024; Duncan, 2023; Gültekin-Várkonyi, 2024; Nebbiai, 2022; Zygmuntowski et al., 2021). It is increasingly recognised as a key development in social innovation (Gegenhuber et al., 2023; Qureshi et al., 2021) and ‘twin transition’ (Horn & Felt, 2025; Leoni & Bartolomucci, 2025; Van De Sanden & Jansen, 2025). The question is not only whether data should be shared, but how, under which rules, and who gets to design those rules. While the European Union continues to pursue innovation through data sharing initiatives (Coche et al., 2024, p. 2), translating these ambitions into practice remains politically and institutionally fraught (Nebbiai, 2022). Recent regulatory initiatives – from the General Data Protection Regulation (GDPR) and Data Governance Act (DGA) to the implementation of sectoral data spaces – introduce new legal mechanisms to facilitate data access, reuse, and protection. Among these, informal regulatory arrangements like industry-specific codes of conduct have emerged as experimental means for codifying market-operator expertise (Divissenko, 2025).

These developments take shape against a background of growing critique of data extractivism (Sadowski, 2019), which questions the capitalist nature of prevailing data economies. In response, various actors have advanced alternative approaches to data governance that emphasise collective benefit and democratic control over data infrastructures. Within the EU, this agenda is reflected in the DGA’s notion of data altruism, which seeks to consolidate voluntary data sharing for public interest purposes under a single legal framework. However, critics such as Veil (2021, p. 4) argue that state-led frameworks, such as data altruism, risk bureaucratizing longstanding, bottom-up practices that already exist under other labels, such as data solidarity (cf. Ansah & Susha, 2024).

Data solidarity refers to the practice of making data processes visible and accessible for public benefit (Bunz & Vrikki, 2022). It emphasizes shared values and mutual obligations among parties involved in data sharing for collective good, particularly in contexts like research biobanks and healthcare (Bak et al., 2023; Braun & Hummel, 2022; Prainsack & Buyx, 2013, 2016). Importantly, data solidarity explicitly challenges hierarchical data structures and neoliberal market forces by emphasising democratic norms and collective benefit (Bunz & Vrikki, 2022). This raises the central question of our study: How and to what extent do co-regulatory sectoral codes of conduct institutionalise solidarity-based data sharing expertise?

We address this through the instructive case of the EU-mandated Code of Conduct for data sharing in the social economy (the CoC), drafted since late 2023 and publicly launched in Brussels in November 2024. The CoC outlines values, practical guidelines, and good practices for social economy organisations characterised by (i) the primacy of social or environmental purpose over profit, (ii) the reinvestment of surpluses in collective or general interest goals, and (iii) democratic or participatory governance (CoC, 2024, p. 4). In addressing organisations that already distance themselves from profit-driven logics, the CoC seeks to codify data sharing expertise aligned with solidarity-based values.

The CoC’s drafting process was distinctive: unlike similar sectoral codes, the CoC was co-produced primarily by civil society actors (including participation of the first three authors of this paper) and with the support, but not the authorship, of the European Commission. Firms were absent from the process, making it a rare case of state-civil society co-regulation. Gorwa (2019) found no state-civil-society-led schemes in his mapping of online content governance initiatives, and Abbott and Snidal (2010) identified only two such cases among 58 regulatory schemes. More broadly, Divissenko (2025) notes the unusualness of civil society leadership in EU digital co-regulation. This uncommon actor mix offers a valuable opportunity to analyse how solidarity-based expertise is claimed, codified, and potentially institutionalised in EU digital governance.

This paper makes two scholarly contributions. Empirically, it provides a situated account of EU co-regulation, based on the participation in the drafting of the CoC of the first three authors of this paper. Conceptually, it extends understudied digital co-regulation research by examining how state and civil-society actors organise themselves to co-produce institutionalisable expertise. Our findings suggest that while the Commission’s facilitation of the CoC’s drafting provided a platform for social economy actors to articulate governance norms, the institutional effects of such instruments ultimately depend on the Commission’s formal approval (e.g. compliance with EU rules, as in mHealth) and symbolic endorsement (e.g. logos, forewords).

We structure the paper as follows: Section 2 outlines the conceptual framework of institutionalising expertise. Section 3 details our methodological approach, combining process research methods, archival work and participant observation. Thereafter, we present our empirical findings. Section 5 synthesises our insights and reflects on the implications for future solidarity-oriented data governance initiatives in EU data policy. Finally, we answer the research question, teasing out implications for the institutionalisation of solidarity-based data sharing.

Section 2 | Organising (for) social economy expertise

As indicated in the introduction, the drafting of the CoC was, at its core, an institutionalisation process centred on the production and circulation of expertise. The consolidation of solidarity-based data sharing within EU governance is not simply a matter of articulating new norms but of embedding particular forms of expert knowledge: who is recognised as an expert, how their claims gain authority and through which organisational arrangements these claims stabilise over time. We therefore conceptualise the institutionalisation process through insights on modes of institutionalising expertise (2.1) on the dynamics of generating, grading and ghosting expertise (2.2), and on the relational, networked co-production of expertise (2.3). Together, these perspectives allow us to grasp how the CoC hovers between a “dead letter” and a “capital good”.

2.1 Institutionalising expertise: Abbott’s three modes

Abbott (1991) suggests that the “proper first question” is: “How is expertise institutionalized in society?” To address how the CoC may advance data solidarity expertise, we take up his distinction between three modes of institutionalisation: (1) commodities, (2) organisations, and (3) professions. Commodified expertise includes forms, manuals, software or any standardised “expert systems” that embed specialised knowledge into a replicable artifact. Organisational expertise resides in the routines and structures of institutions (firms, agencies, cooperatives, state bodies) where knowledge is encoded in roles and procedures. Professions, by contrast, centre expertise in credentialed practitioners and their guild-like controls.

The CoC can be read as both commodified and organisational expertise. As a text with principles, templates, and checklists, it is itself an expert commodity: a reproducible package of specialised knowledge about data sharing produced by so-called social economy actors. Anyone can download and reuse this “code”, much like a handbook or open database, applying its logic to real data-sharing projects. At the same time, the CoC emerged from a temporary drafting committee and only acquires real force if adopted and enacted within organisations. In other words, the Code is an expert commodity shaped by (and for) an expert organisation. Professions, by contrast, were marginal to the process, since the production of the Code relies less on accredited experts than on civil society actors and the Commission.

Moreover, Abbott (1991) emphasises that these modes are complementary: the CoC gains traction and is made effective when translated into the routines and procedures of social economy organisations. These organisations, in turn, can gain legitimacy by demonstrating adherence to the CoC as an official, authoritative text. The fragility of the CoC rests on its failure so far (although, it is still early days) to transition from portable “capital good” (ibid., pp. 22-23) to becoming ingrained in everyday data sharing practices within the EU social economy. As is, the CoC risks remaining a dead letter until its principles and templates are routinised.

2.2 Generating, grading, ghosting expertise

If Abbott (1991) clarifies where expertise is institutionalised, Monteiro (2024) directs attention to how organisational structures actively shape what expertise becomes. This distinction helps us understand the CoC’s institutional context. Monteiro (2024) argues that organisations both codify knowledge in formal artefacts (e.g. handbooks or standards) and define the domains within which experts can legitimately speak. In the case of the CoC, two structures were particularly decisive: first, the temporary drafting arrangement, comprising the Contractor, Project Partners, selected experts and Commission oversight (see Section 4 for further elaboration), and second, the CoC text itself, which organised values, guidance and good practices into a structured classification system. Together, these structures defined the boundaries of what counted as social economy expertise on data sharing.

Monteiro (2024) identifies three heuristic processes through which such boundaries take shape: generating, grading and ghosting. In the CoC, the drafting committee generated expertise by defining new categories (e.g. list of values such as “trust” and “self-determination”), graded it by modifying expert statements (e.g. specifying “data sufficiency” rather than only “sufficiency”), and ghosted expertise by excluding certain expert statements (e.g. dropping “innovation” as a value). These processes signal shifting judgments within the temporary drafting arrangement about what kinds of expertise belong within the solidarity-based framing and which do not.

The operationalisation of these heuristics anchors our empirical coding strategy (3.2), allowing us to trace how solidarity-based expertise was claimed, codified and contested throughout the CoC drafting process.

2.3 Expertise: a relational lens

In applying Monteiro’s heuristics, we do not treat the CoC as merely the sum of Abbott’s modes of institutionalising expertise, but as a relational capacity (Emirbayer, 1997; Eyal, 2013; Pakarinen & Huising, 2023). The CoC’s expertise is thus an emergent property of a heterogeneous arrangement, including the understudied state-civil society actor mix (see Section 1), regulatory standards, and policy concepts that are mandated by the EU to address its social economy’s data sharing concerns. From this relational standpoint, the CoC becomes “alive” when its concepts travel across sites, are referenced in funding calls, or embedded into social economy projects. Conversely, it becomes a “dead letter” when these connections stall and the CoC ceases to mobilise actors or resources. In this way, we analyse how the CoC assembles values, procedures and devices (e.g., “data sufficiency”, non-extractive models, data sharing agreement templates) that collectively enable data sharing. In practical terms, studying the CoC means tracing how expert statements change throughout the drafting process with the goal of understanding how solidarity-based data sharing expertise is dynamically “claimed, recognised and institutionalised vis-à-vis other forms of expertise” (Pakarinen & Huising, 2023, p. 9).

We focus on values, procedures and devices, because they function as “immutable and combinable mobiles” (Eyal, 2013; Latour 1987, p. 227). Devices and concepts allow an expert statement to be transported from one site to another without deformation and gain the traction, which is necessary for the commodity expertise to be shared with and used widely by policymakers and practitioners. To illustrate, consider the CoC’s dissemination via the official Commission website. This communication channel extends the reach of the CoC. Seen relationally, Eyal (2013) describes the power of expertise, such as the CoC, as “generosity” (i.e., extending concepts and methods to others to elicit cooperation) and “co-production” (i.e., involving multiple parties in shaping expert knowledge).

Generosity can empower a network by having the “capacity to craft and package its concepts… so they can be grafted onto what others are doing, thus linking them to the network and eliciting their cooperation” (Eyal, 2013, pp. 875‒876). The CoC itself aims to be accessible to those without deep specialised training, thereby exhibiting generosity to elicit cooperation among social economy actors and beyond. Furthermore, co-production refers to the situation where a network gains influence by “involving multiple parties… in shaping the aims and development of expert knowledge” (ibid., p. 876). The CoC drafting was labelled a “co-creation process” by the Commission, involving 21 experts from Member States and two facilitating organisations. Thus, from the sociology of expertise perspective, the mandate for the CoC drafting process called for the Commission to cultivate a “generous” temporary organisation for its “co-production”.

Section 3 | Methodology

This study is qualitative and process-oriented. We combine archival research, participant observation, and process analysis to investigate the EU-mandated drafting of the CoC. Our analysis is sensitised by Monteiro's (2024) heuristics of generating, grading and ghosting, and by the network turn in the sociology of expertise (Eyal, 2013): we trace how expertise was temporarily stabilised in the iterative successive drafts produced through interactions between policymakers, civil society organisations, social economy actors, and domain experts (See Figure 1 for actor constellation diagram). Process analysis (Langley, 1999) is appropriate because the CoC is an iterative document, with each intermediate draft serving as an empirical artefact of expertise-in-the-making.

3.1 | Data collection

Our empirical material combines three sources: archived documents (see Annex II), observational material (see Annex II) and reflexive discussions among the authors, produced through our layered involvement in the drafting of the CoC. Two co-authors participated in the drafting from its inception in late 2023, explicitly motivated by a commitment to institutionalising data commons, a specific variant of digital solidarity economies, within EU governance frameworks. The lead author joined in April 2024 as part of his doctoral project, positioned at once as contributor to the process and as its analyst. The final co-author entered only after the launch of the CoC, assuming the role of critical discussant and supervisor, providing methodological oversight and analytical distance from the field.

While some of us acted as active members (Adler & Adler, 1987), directly shaping the CoC’s content and direction towards our normative ideals in varying degrees, the final co-author remained closer to peripheral membership observing (ibid.), commenting and interpreting as critical readers of both the process and its outcomes. Our collaboration therefore embodies a neither fully immersed nor fully detached “insider-outsider perspective” (Dwyer & Buckle, 2009).

This mixed perspective sharpened our analysis. Insider status afforded depth of access (Kanuha, 2000), allowing us to capture discussions and dynamics that cannot be observed from the outside, such as old drafts, attendance to the technical sessions (disclosed from the public minutes) and internal communications (e.g., chat logs, working documents, emails). However, we also confronted the risks of role conflict and blind spots (Brannick & Coghlan, 2007). To mitigate these tensions, we triangulated between archival material, documented internal communication, and collective reflexive dialogue.

3.2 | Analysing content changes in the drafting process

We treat the CoC as an iterative artefact of commodity-expertise-in-the-making. Our coding proceeded in three iterative cycles.

In the first cycle, we tagged edits by draft phase for temporal scaffolding. In the second cycle, we cross-referenced these tags with meeting minutes, communications, and field notes to explain how specific edits were linked to events. In the third cycle, we coded in the iterative draft versions manually for Monteiro’s categories, as delineated in Table 1 below, marking whether expertise statements were introduced, changed or deleted. The outcome is a processual account of how expertise was materially inscribed, reworked, and erased across the drafting process.

For narrative clarity and with “the goal of being read” (Czarniawska, 2008, p. 14), we dramatise the drafting process in three acts. These acts correspond with observable shifts in empirical data and actors involved: Act I covers publicly minuted sessions; Act II examines internal drafting work; Act III follows the Commission’s interventions that shaped the final text. The empirical anchors throughout the phases are the iterative drafts and the generation, grading, and ghosting processes we identified through our analysis. We include a visual map (Langley, 1999), linking the three expertise processes to the main actor clusters and to the corresponding CoC draft chapters.

3.3 Research context

Finally, to situate the CoC’s drafting process, we trace its policy genealogy. It is rooted in the EU’s 2020 industrial strategy, which initially identified the sector “Proximity, Social Economy and Civil Security” (SWD 351, 2021). This sector’s name has by far the most characters (44)¹of any industrial sector. Yet this breadth proved short-lived: in later revisions, “civil security” was sidelined on the ground of “minimal economic activity” and its close ties to public services (Transition pathway for proximity and social economy ecosystem, p. 8). Furthermore, the “proximity economy” was gradually folded into the social economy, with the European Commission (EC) framing this field as the ecosystem’s “centre of gravity.” (ibid.). Officially administrative and pragmatic, these semantic shifts also have policy consequences: institutional actors use language to stabilise which organisations, activities, and values count as part of the EU’s “social economy” and, therefore, who is recognised as an expert and who is invited to the table.

Such reframing is not unique to Brussels. Across Europe, the notion of social economy has been shaped by diverse legal and institutional traditions, and alternatively labelled as, say, cooperatives or commons (Yi, 2023). In other contexts, “the term ‘social economy’ is substituted for other terms, such as ‘Social and Solidarity Economy’, ‘Cooperative economy’ and ‘Third sector’” (CoC, 2024, pp. 4‒5). In EU usage, the social economy is domesticated as a policy sector, aligned with industrial strategy, yet its political ontology remains relational, emerging in friction with dominant regimes (Longhurst et al., 2016).

By contrast, digital solidarity economies offer a more expansive, extra-institutional concept. They encompass collective practices of organising, sharing or governing digital resources through non-extractive, mutualistic, and democratic logics. Unlike the formalised EU social economy, they emerge outside of institutional classifications. Yet they circulate across activist, scholarly and policy domains.

Thus, the CoC sits at the intersection of bureaucratic narrowing and solidarity expansion. It is a hybrid project, designed to render diverse data governance practices legible to EU industrial policy while simultaneously articulating a values consensus (democracy, self-determination, inclusivity) aligned with solidarity imaginaries. In this tension, the CoC can be read as an effort by social economy actors to claim institutional space for alternative governance logics, sometimes in friction with EU rules, sometimes in hybridisation.

In short, we treat “social economy” and “solidarity economy” as different but closely affiliated labels mobilised strategically in the European context. Based on the briefly sketched genealogy above, we recognise a certain administrative, if not political, pragmatism in the Commission’s categorisation practices. Other work (e.g. Longhurst et al., 2016; Parnell, 2021) does distinguish ontological differences between these and more diverse economies. Therefore, we invite readers to engage with our use of digital solidarity economies as descriptive and plural. We explicitly theorise how such alternative expertise may (or may not) be institutionalised in EU data governance.

Section 4 | Staging expertise: generating, grading, and ghosting

The CoC was drafted over twelve months through a staged co-creation process convened by the European Commission’s Directorate-General for Internal Market Industry, Entrepreneurship and SMEs²(hereafter: EC or the Commission). The project was executed by a contractor consortium led by Waag Futurelab³(the Contractor) and Commons Network⁴(Project partners, also the first three authors of this paper). The timeline began with an open call in late 2023 and culminated in the public launch in November 2024.

From this call, the Commission selected 21 “frontrunner” experts representing organisations active across Europe’s broadly defined social economy. Their fields of expertise ranged from open-data advocacy and cooperative governance to platform-based social entrepreneurship. While many participants shared affinities with digital solidarity economies, their institutional affiliations remained diverse, spanning cooperative federations, social innovation networks, civic tech organisations, research institutions, and even one municipal body. To structure inputs provided during the drafting process, the Contractor divided the experts into four roughly even-sized subgroups, visualised in Figure 1 as drafting groups 1‒4.

The Commission acted as convenor, issuing the call, hosting the kick-off meeting and overseeing the process. Meanwhile, the Contractor coordinated logistics, designed the co-creation framework and took responsibility for producing Chapters 1 (articulating the social economy’s data-sharing related values) and 3 (inventory of good data-sharing practices). As Project partners, we contributed structured feedback on the Contractor’s output and took responsibility for Chapter 2 (guidance for social economy actors), including ready-to-use data sharing agreement templates for social economy actors. Finally, the Special Group of 21 experts served as the project’s deliberative core: co-writing, reviewing, and sharing examples of “good data sharing”.

The process unfolded through four publicly documented co-creative “technical sessions” (two in-person: Brussels, Amsterdam; two online), which chronicle and corroborate the CoC as the product of collective authorship. In practice, however, the bulk of textual drafting occurred backstage between September and November 2024, as narrated in Acts II and III, through asynchronous exchanges, bilateral revisions, and even an impromptu focus session. Although the CoC’s final report describes its participatory method, the behind-the-scenes process of how expertise was negotiated and staged remains largely invisible. Our analysis demystifies this process by drawing on our involvement as project partners to reconstruct how expertise was assembled, ranked, and sometimes silenced through three distinct yet overlapping processes: generating, grading, and ghosting (Figure 2).

In short, the drafting of the CoC took place at the intersection of evolving EU policy categories, the emerging extra-institutional discourse on digital solidarity, and diverse normative ‘expert’ visions of the social economy. Bringing together actors ranging from cooperative networks to EU civil servants, the process sought to define good practices for data governance in the sectoralised social economy.

Act 1: Opening stage

On 12 July, the Contractor shared the first draft of Chapter 1 (v1.0) with the four group chairs and the EC representative. The draft was uploaded to a shared Nextcloud folder and distributed via email. In the accompanying message, the Contractor emphasised that the text drew on the third technical session and had been modelled after existing codes of conduct in other data spaces, namely that of agriculture, tourism, and mobile health apps (mHealth) (see Annex I).

The structure of the chapter followed a conventional format: an introduction describing the data space of the social economy, followed by a list of eight value categories. The Contractor acknowledged that they had “made choices” in merging the results from the previous technical sessions and asked the chairs and EC to review the text with three guiding questions: were the most crucial values included? Should the descriptions be revised? Should any values be omitted or, if necessary, added? While feedback was invited, the message made clear that the preference was for reduction rather than expansion: “we would rather reduce that number, not enlarge it“ and avoid “re-opening the discussion and start all over again”. At the same time, space was left for “something fundamental” to be raised, which could be discussed in the September meeting. A deadline for comments was set for 26 August, six days before the final technical session.

By the last technical session, the Contractor had compiled the first round of feedback. Discussion focused on clarifying scope, balancing the format and deciding on which values to include. According to the public minutes, several content changes were made to the code of conduct. First, committee members stressed that the CoC should be limited to values guiding data sharing in the social economy, rather than addressing broader challenges such as funding or digital skills. Second, the group agreed to adopt a more uniform structure: starting each subsection with a question and ensuring that each value was explained in text of roughly equal length. Third, a new value of self-determination was added, to shorten the extensive section on trust and redistribute references to sub-values autonomy and transparency. Clarifications were also requested for equity/fair distribution of value, which participants found vague.

Other comments concerned implementation of the CoC. Committee members asked how the CoC would be monitored or assessed, whether benchmarks or metrics would be used, how effectiveness would be evaluated, and how the code would be disseminated and in which European languages. The Contractor acknowledged these as important, but beyond the project’s scope and depend on follow-up initiatives by the Commission.

The second part of the final session was an update on the collection of best practices. If the first Chapter had been about reduction and stabilisation, this chapter was about expansion. The Contractor outlined a course of action: Chapter 3 of the CoC would contain at least 30 examples of data sharing, each framed as a “good practice” By email, the Contractor shared the first version of the good practices template, inviting EC and Project partners to comment. The EC official refrained from any input, while we, as Project partners, intervened to expand the form with questions around licensing, consent, and data management. Contributions before the final session, however, remained uneven: group 1 submitted three templates; group 2, six; group 3, seven; and group 4, two. During the session, the Contractor positioned themselves as curator: they would “review all the provided good practices and complete the information.” Here, expertise was being generated and graded, but also commodified into a bureaucratically manageable form. The Facilitators asked for more template submissions by committee members to reach their target (30).

The remainder of the last technical session was more open-ended. Our team facilitated a workshop on governance models for data intermediaries, using break-out groups and plenary discussions. Participants were asked to consider intermediaries’ roles across different domains (food, energy, health, and mobility). Unlike the highly structured exercises on values, this part of the workshop produced no direct textual outcome for the CoC, instead grading provisional data sharing models identified through desk research and interviews.

One of the clearest examples of downgrading, and eventual ghosting (see Act 2), emerged around the value of “innovation”. During the third session, the Contractor facilitated a workshop in which the committee members were asked to articulate and deliberate on social economy values⁵. The public minutes record how the inclusion of “innovation” was immediately questioned. An unnamed participant raised the ambivalence of treating innovation as a value in itself:

“You mentioned the value of innovation and its direction, also highlighting the possibility of ending up with a worse outcome or innovating without a clear incentive. You also talked about openness to new ideas and social innovation. Have you considered alternative terminology for this concept?”⁶

This ambivalence persisted into the written review process. Group 2 was the only drafting group to return to the issue explicitly through comments placed in the Nextcloud-stored document. The feedback oscillated between efforts to reframe innovation as compatible with social economy by changing the subsection title to “social innovation” or “innovation for the common good.” One member suggested elevating innovation as a transversal principle, comparable to democracy, and even strengthening it:

“SE actors should perceive data as a strong innovation trigger and approach the data use as an occasion to not only better describe social phenomena but also as a way to re-design their services and products.”⁷

Taken together, Act 1 illustrates how organisational structures shaped the drafting work. Deadlines, templates, and workshop arrangements narrowed the range of possible inputs, while the contractor’s instructions framed what counted as feasible revision. Thus, the first draft was progressively reshaped into a format deemed acceptable to the committee and aligned with broader expectations for EU codes of conduct.

Act 2: Behind the curtains

If Act 1 played out in public, Act 2 unfolded backstage. After the last technical session, the drafting process shifted from collective discussion to a flurry of document versions, email exchanges and a tightly managed online focus group. This was the phase of consolidation: carrying the tentative outputs of Act 1 toward a pre-final draft that could be sent to the Commission.

The proliferation of draft versions (v1.1 on 4 October; v3.0 on 16 October; v2.0 on 17 October; v1.2 and v2.1 on 3 November) signalled intense editorial work. Iteration here was no longer about collecting input, but attention shifted to testing coherence across chapters, aligning terminology and tightening scope. What emerged was a pre-final draft ready for the decisive interventions of Act 3.

Chapter 1

The development of Chapter 1’s values can be traced across versions (v1.0–v1.3). Rather than line-by-line edits, the focus here is on how values were generated, graded, or ghosted during this act. Table 2 summarises the changes in value categories.

In the initial draft (v1.0), “trust” bundled sub-values (e.g. autonomy, consent, transparency), while “innovation” appeared as a contested value. By v1.1, this terrain had shifted. The drafting committee generated a discrete value of “self-determination”, relieving trust of its conceptual overload and creating space to treat autonomy and consent more explicitly. At the same time, the committee graded several compound categories into leaner terms: “empowerment and inclusiveness” was streamlined to “inclusiveness”, and “fairness/equitable distribution of value” was reframed as “equity”. Most strikingly, “innovation” was ghosted, abandoned altogether. What looked like an open debate on grading the value category, in fact shaped its deletion from this document. This deletion shows how a value can be downgraded out of the final draft.

The erasure of “innovation” and the accentuation of “self-determination” illustrate the contest over what counts as solidarity-based expertise. This contestation reveals how solidarity values were not uniformly strengthened, with some (e.g. equity, self-determination) institutionalised as legitimate moral anchors.

By v1.2 values stabilised, with only minor refinements in v1.3: “inclusiveness” was relabelled “inclusivity” and “sufficiency” was sharpened to “data sufficiency,” thus anchoring a normative claim in more technical terms. These late tweaks signal grading at work, with the committee clarifying which values define data sharing in the EU social economy and qualifying them technically when necessary.

Seen through our heuristic framework, two patterns stand out. First, generation is visible in the responsive addition of “self-determination”, an example of organisational problem-solving that creates analytic space for contested sub-values. Second, grading operates as continual pruning and specification: broad, compound claims are narrowed to categories that are easier to operationalise. Third, and most notable, the progressive ghosting through downgrading of “innovation” marks a political as well as semantic choice. Removing innovation at v1.1 displaced an EU industrial axiom that treats innovation as an end in itself. Innovation was demoted and relevant expertise statements absorbed by “co-opetition.”

Chapter 2

In parallel, project partners cultivated Chapter 2. An online focus group (18 October) combined presentations with controlled solicitation of feedback. The agenda divided attention across three dominant themes in Chapter 2: data intermediaries, data sharing templates, and policy recommendations. Each segment was organised around guiding questions: (1) “what is missing?”; (2) “what should be omitted?”; and (3) “what sector-specific terms should we include?” While signalling openness, these questions also narrowed the scope of acceptable feedback.

From this session, two notable generating moves followed. First, drafting committee members pushed for the inclusion of “data collaboratives” as a model. This was remarkable precisely because it contradicted the Commission’s institutional framing: the 2024 JRC report, from which the CoC draws heavily, had excluded data collaboratives “because they cover a wide range of data relations which do not conveniently fit the European framework of data intermediaries” (final CoC, 2024, p. 25). Their addition thus reasserted solidarity-oriented logics against institutional classification. Second, project partners generated a typology of extractive and non-extractive data relations (ibid, p. 23), explicitly operationalising values of equity (compensation) and self-determination (consent). This second move also implicitly upgrades stewardship logics while downgrading extractive ones, aligning explicitly with data solidarity.

Grading was not limited to feedback triggered by the EC or the drafting group. It also emerged in the project partners’ internal Signal exchanges, where self-reflexive discussions prompted the refinement of the conceptual schema underpinning the CoC’s guidance section (Chapter 2). In a November 20 Signal thread, partners reconsidered the use of the label “honeypot” to describe data-sharing arrangements offering fair compensation but based on meaningless or coerced consent. Recognising that “honeypot” carried an incompatible cybersecurity connotation, we opted to relabel the category as “transactional logic”. Rather than discarding the underlying idea, we relabelled it “transactional logic”. In effect, we graded our own schema for communicability and fit, demonstrating that evaluative narrowing could emerge just as much from reflexive sense-checking as from Commission or Special Group feedback.

Alongside this, the Project partners generated the first draft of data sharing agreement templates, which offer practical steps for establishing non-extractive data relations. Developed from earlier checklists, the Project partners had now written it in contractual terms: “why not just create the language that could be used as a CoC?”⁸, a Project partner asked rhetorically in the accompanying email. These templates turned abstract principles into reproducible “immutable mobiles” (see Section 2.3) that could travel across contexts. Here, generation and commodification converged: expertise was packaged as portable templates. Furthermore, the Contractor graded these drafts by pressing for stronger links to Chapter 1’s values. For assurance, the templates also underwent external review by a pro bono legal advisor, sourced from the Project partners, not the EC.

Chapter 3

On 16 October, the Contractor first shared Chapter 3’s catalogue of good practices. Initially 33, these were curated down to 30. The standardised template provided by the Contractor functioned as a generative device: it transformed disparate submissions into comparable “cases” that could be slotted into the catalogue.

Not all contributions originated with committee members. Some were harvested from the technical session outputs (e.g. canvas boards, chat logs), others recycled from an earlier self-regulatory attempt by an organisation affiliated with Group 4, and still others submitted by individuals with no clear institutional affiliation⁹. Through this aggregation and formatted work, the Contractor generated an Excel overview that could be presented as the collective knowledge of the drafting group.

After generation came grading through curation: Of the 33 initial cases, four were downgraded as “more information needed,” and one flagged as “optional”. Grading also occurred in subtler ways: minimal feedback from the project partners (“I think we can get more out of the template”), signalling compliance pressure with standardised forms, where “more out of” implicitly means greater compliance with the standardised structure. By November 1, the Contractor acknowledged by email the limits of grading: the requested information on licenses and data models was too sparse to allow for deeper comparative assessment and the facilitators are “out of time”. Instead, “value-analysis” became the main method to make sense of the good practices. Here, grading becomes both narrow and self-limiting, producing a chapter that could only narrate alignment with abstract values rather than infrastructural and legal arrangements.

Version 3.1 consisted of 30 entries and an introduction. Three entries were ghosted, yet traces of one (Mobicoop) lingered in the final draft as a reference point for data cooperatives (final draft CoC, pp. 27‒28). The introduction framed the result as stemming “from the manifold experiences of people working in the social economy across the EU and beyond.”¹⁰This highlights the curation process as a distributed, participatory knowledge base, even where provenance was patchy.

By 5 November, the Contractor assembled all three chapters into the first full draft of the CoC and shared it with the committee members for last-minute feedback before sharing the text with the EC. The next day, the text was shared with the Commission. The archive does not detail feedback received from the committee members in that 1-day window other than one comment with minor text edits.

After sharing the pre-final text with the EC the next day, the Contractor’s and project partner’s attention shifted to organising the launch of the CoC, Brussels 28 November, until further notice from the Commission. At that moment, editorial authority shifted away from the drafting group toward the EC.

Act 3: Final cut

The third act opens with EC’s mid-November feedback. While their presence had been consistent throughout earlier phases (e.g. hosting technical sessions 1 and 2), the final act made their influence more visible and also more contradictory. On 15 November, the Commission provided feedback on all three chapters of the draft. The comments were extensive, ranging from editorial suggestions (e.g. lay-out, offering examples under values) to structural interventions (e.g. reorganising annexes, inserting new concepts and a good practice). The email chain captures the key tension of this act. On the one hand, the EC introduced new concepts (e.g., tech neutrality and decentralised autonomous organisations) and pressed for greater editorial emphasis on issues like transparency and tech neutrality. On the other, they carefully distanced themselves from the CoC by denying use of their logo and inserting a disclaimer disavowing authorship. In other words, the Commission convened, facilitated, and substantively shaped the process, yet insisted that the final text should not be seen as their official position.

On a mid-November Friday, the EC’s consolidated review arrived. The message praised overall readability but flagged substantive gaps: Chapter 1 “misses very short examples on how a value could lead to an action,” and Chapter 2 “is not bridging well to the [data sharing agreement] templates” and “is still very meta level for a guidance.” Even the geography of the collected good practices in Chapter 3 was opened for revision: “we only have 10 MS [Member States] covered.” The email was long, detailed, and unmistakably directive: “Ideally we cover at least 50% of the MS,” “DAO is missing and should certainly be added and explained,” “Maybe the templates… could follow after this chapter.” The email closed with reassurance: “it looks like a lot of comments, but… it’s mostly quick fixes”, which seemed to soften the extent of intervention.

The response from us as Project partners, sent later that evening, accepted the spirit of the comments while pushing back and asking for clarification. “Overall good points, we will try to integrate as much as possible,” our email opened diplomatically. Questions quickly followed: “what aspect of tech neutrality do you find important to highlight? as it is a normative concept which actually one could also question”; “what would you suggest for the best practices then?” The email politely but firmly asked for clarification, signalling two practical limits. First, as drafters we did not want to simply rubber-stamp everything. Second, time was running out, with less than two weeks before the launch event in Brussels, integrating conceptual changes seemed unfeasible.

The following Monday, the Commission remained managerial and prescriptive in its response and reframed the task with a liberal dose of user-centric pragmatism: “Try to think about the reader as having no clue.” They propose concrete fixes (e.g., reintegrate the checklist immediately after Chapter 2 and relegate templates to the annexes). The EC reasserts that tech neutrality must be fully explained and linked to practical guidance so readers can assess how technology choices map to values.

Between these exchanges and the final draft, the Contractor and Project partners reworked chapter links and moved material (checklists, templates, good practices), went back and forth on newly proposed concepts (e.g. tech neutrality, DAOs) and incorporated an extra practice suggested by the EC: “I added one case myself from Germany (MS extra) – this was discussed with [the Contractor]. Please have a look if acceptable for you and to link it to the appropriate values. I suggested several ones.”

An instance of late-stage generation, and subsequent downgrading, was the introduction of decentralised autonomous organisations (DAOs). The passage below presents a cautious acknowledgment: DAOs are framed as a possible future model for data governance but, at the same time, the CoC distances them from the social economy by questioning their reliance on smart contracts rather than trustful human relationships, a core value established in Chapter 1. DAOs are positioned as incompatible with the ethos of solidarity and trust central to the social economy. Thus, DAO’s inclusion can be read as a performative gesture, offering the appearance of forward-looking while ultimately reinforcing the boundaries of accepted models within the social economy’s imagined EU data economy. The notion of tech neutrality failed to generate in the CoC.

“Decentralised autonomous organisations, also known as DAOs, can also be a model for democratic data governance over data as shared digital assets. DAOs rely on smart contracts, which are self-executing computer programs used to enact decisions that automatically go into effect once certain conditions are met. There are some DAOs that are experimenting with governing data as a shared asset, however due to their reliance on complex socio-technical infrastructure and ambiguity around their legal status and liabilities there is still significant barriers to their widespread adoption. It is not obvious that DAOs are in line with social economy values, as smart contracts are conceived of as a way to do away with the need for trustful relationships”

(CoC, 2024, p. 28; emphasis ours)

One of the most striking textual interventions, which was not mentioned in the accompanying mail, was DG GROW’s addition of the disclaimer. This disclaimer complicates our heuristic analysis. It can be read as generating (a new text), but also as ghosting (erasing the EC’s own role in shaping the CoC). Throughout this act, the Commission served as a ghosted fifth drafting group: generating concepts and examples, grading acceptable framings and ghosting their own imprint.

On 26 November, the CoC was finalised by a graphic designer and circulated in pdf. Despite the Commission’s close organisational involvement, the final document does not represent an official European Commission position (see Figure 3 for disclaimer in the final draft). The circulation mail from the contractor was upbeat: the CoC would go live on the Commission’s website in two days, with a social media promotion to coincide with the Brussels launch. Yet compared to the code developed in the tourism sector, whose circulation was buttressed by an EC foreword (see Figure 4), the social economy CoC entered the world with weaker symbolic backing. Within half an hour, a project partner raised concerns about its dead-letter status:

“I think it's important to at least include the Commission logo, so it's a bit more official? Even if it's just as a funder/client. Right now, there's no clear sender when you take a quick look. (…) I really think it's very important for the adoption and seriousness of the code.”

(authors’ translation from Dutch)

The request was denied within four minutes. The Project partners pressed again. The Commission clarified that logo use was prohibited in this case because the CoC was not a formal deliverable meeting the Commission’s template and approval “against pre-set standards.”¹¹This ambivalence is political. By intervening substantively while disclaiming responsibility, the EC shaped the boundaries of legitimate ‘social economy data sharing expertise, without being held accountable for the text itself. The refusal to allow logo use made this clearer.

Two days later, the official launch of the final draft was held at a Commission venue in Brussels. At the event, the CoC was presented as the outcome of a collaborative social economy process, while the Commission’s fingerprints remained everywhere, except in the formal markers of endorsement.

Section 5 | Discussion: expertise in process, four lessons learned

Our insider analysis of the CoC’s drafting reveals collaborative norm-setting as a site of institutional work. This expertise was coaxed into being through a relational interplay of actors, tools, and ideas. This process empowered social economy actors by articulating values (Ch. 1 of the CoC) and introducing new norms (Ch. 2 and Ch. 3). Following Monteiro’s (2024) heuristics, we traced how statements were generated, graded, and ghosted, and how these patterned expertise processes were staged across three acts. Abbott’s (1991) question of how expertise is institutionalised, and Eyal’s (2013) emphasis on generosity and co-production, guide our reflections here. Drawing also on our role as facilitators of the drafting process, we distill four lessons learned.

First, unpacking the drafting through generating, grading, and ghosting allows us to shed light on how substantive elements of the CoC emerged. Value categories such as “self-determination” came together in the figurative drafting room, while tropes like “innovation” were ghosted out. Such moves demarcate the boundaries of the CoC, shaping which solidaristic values travel in an EU context. At the risk of self-aggrandising, the Facilitators’ role was crucial here: winning the tender bid meant convening, translating, and editing a “generous” temporary organisation for the “co-production” of solidarity expertise (Eyal, 2023). We show how the CoC is provisionally curated and staged.

Second, the EU’s reliance on voluntary cooperation raises serious questions about the adequacy of its support. The emergence of commodified expertise within the CoC is not an inevitable outcome of spontaneous actor-interactions. Rather, it is the result of concentrated effort, sustained coordination and significant paid and unpaid investment. In this sense, the CoC is a collaborative achievement situated within the reign of volunteerism, where policy development depends on institutional work performed by informal actors without remuneration or mandate. Facilitators were allocated funding, whereas social economy experts received no financial compensation despite contributing extensive time and effort in technical sessions and document review, illustrating Eyal’s (2013) generosity.

Third, in the soft law context of codes of conduct, symbolic markers matter. The Commission facilitated the process and was given the last round of feedback, yet withheld formal endorsement. Curiously, our analysis leads us to wonder, perhaps aggrievedly, whether requesting a foreword from a high-ranking European Commission, rather than the in hindsight unattainable EU logo, might have anchored greater legitimacy, as in the tourism code. Such seemingly banal symbols mask the deeper mechanisms by which expertise becomes policy-relevant. Without adequate process expertise, facilitators (including ourselves) may miss opportunities to secure recognition for the CoC. At the same time, the official launch in Brussels and hosting of the CoC on the Commission’s official site raise the question of whether such visibility suffices to validate its expert status.

Fourth, the Commission’s disclaimer and refusal to allow logo use highlight a paradox at the heart of the CoC’s actor-mix. The CoC was publicly presented as a co-creation of social economy actors, yet it ultimately depended on the Commission’s institutional support (e.g. through convening the Special Group, funding the project, hosting formal sessions in Brussels). The paradox thus lies between authorship and ownership: the CoC’s legitimacy derives from its social economy provenance, while its policy relevance depends on proximity to the very EU structures it seeks to complement and sometimes critique. This balancing act produced hypervisibility for some experts (unlike other sectoral codes, the CoC names individual drafters and their social economy affiliations) while ghosting others (e.g. individual Commission advisors and the pro-bono legal reviewer). Whether the denial of the logo represents a temporary hindrance or a structural barrier to the full institutionalisation of the CoC as solidarity-based expertise, remains a matter of future empirical investigation.

Section 6 | Conclusion

How, then, do co-regulatory sectoral codes of conduct institutionalise solidarity-based data sharing expertise? Our analysis of the CoC’s drafting process showed expertise emerging and stabilising through processes of generating, grading, and ghosting by both state and civil-society actors. The CoC is best understood as a fragile institutional asset: a commodity that circulates as a portable text and a set of organisational rules that can be enacted, yet one whose uptake is not guaranteed.

We used Abbott’s (1991) three modes to conceptualise that the CoC institutionalises primarily through commodified and organisational modes. As a freely accessible document, it can circulate as an “immutable mobile” (Latour, 1987), simplifying solidaristic norms into legal templates that can be redeployed by practitioners. As an organisational resource, it is manifestly designed to guide social economy projects under the DGA. However, without embedding into routines, the CoC risks remaining a dead letter.

Eyal’s (2013) concepts of generosity and co-production help us make sense of this. Templates and examples embody generosity, extending knowledge to non-experts and creating a toolkit or repertoire of data solidarity expertise. Co-production is visible in the layered process of drafting: from broad deliberation to consolidation, to Commission officials distancing themselves from the final content through a last-minute disclaimer statement.

Whether the CoC succeeds in institutionalising solidarity expertise depends on future uptake by the EU. In mHealth, failed approval stalled institutional effects; in tourism, their code of conduct gained traction as part of the development of a common data space. The CoC could follow either path, depending on whether policymakers and practitioners reproduce it as a living resource rather than leaving it dormant. Future references to the CoC by EU regulatory bodies such as the European Data Innovation Board, European Data Protection Board or the Data Spaces Support Centre will signal whether it becomes a “capital good” that generates returns for solidarity.

Seen through the lens of digital solidarity economies, the CoC is neither triumph nor failure. It is a provisional achievement, staging data solidarity in ways that validate and empower its expertise within regulatory standards. It also exposes the reliance on unpaid labour, soft-law instruments, and ambiguous institutional support. As such, the CoC embodies a digital solidarity economy dynamic in miniature: collective efforts to create non-extractive alternatives within a broader economy still mired by innovation and technological neutrality tropes. Rather than a static “dead letter”, the CoC can be read as a capital good whose produced value depends on ongoing translation work. If future actors take up its data sharing agreement templates, integrate its principles into organisational routines or use it to negotiate new data-sharing arrangements, the CoC may become a durable node in Europe’s solidarity-based data regime. Conversely, if it remains unadopted and unreferenced, it at least convened and temporarily stabilised a network of diverse actors (e.g., government officials, activists, researchers) to perform organisational and political work to codify ethical codification.

Its future trajectory, whether the CoC remains a fragile artefact or becomes part of a durable data practice, will shape how far solidarity-based expertise can be institutionalised in Europe’s new data regime.

Annex I

A comparative glance at other EU sectoral codes reveals divergent institutional trajectories:

Annex II