Abstract
This article examines how Wikipedians embed their sovereign authority within the development of the site’s multilingual policy environment. By drawing on the concepts of cultural techniques, imagined affordances, and online authority, the edits and comments of editors were examined through a discourse analysis of 15 rules across 15 years. With a focus on the English and Spanish-language Wikipedias and additional comparisons with the Arabic, Dutch, and French-language editions, our analysis presents a new perspective on Wikipedian self-governance. By paying attention to how Wikipedians designate the authority of rules through templates, we observed that an incredibly small number of editors use techniques to designate the sovereign authority of policies in ways that determine community consensus, rather than reflect it. This research provides further understanding of the democratic limitations of online peer production and self-governing communities dedicated to equal and diverse forms of participation.
Keywords
Introduction
Every policy on Wikipedia reflects community consensus. At least that is the dream of self-governance that holds the complex project together. Ensuring that over tens of millions of different users on 326 separate language-editions (Wikipedia, 2024) consistently agree on how to write articles and coordinate their work is a difficult task. But this difficulty has not stopped Wikipedians. Instead of relying on top-down decrees that prescribe user behaviors for both veteran and new users, Wikipedians rely on ‘best practices’ that have been inscribed in the site’s guiding documents (Wikipedia, 2009a). As such, their “diverse equality” as editors “is constrained in one aspect – the set of required attitudes and standards” that transform atomized users into a community of Wikipedians; users who share practices and values associated with producing the encyclopedia (Pentzold, 2011: 714). But how was this diverse equality expressed when the norms of the community had not been formalized? In other words, did the production of Wikipedia’s rules reflect the values that it is so widely known for? We argue that despite good intentions, many of Wikipedia’s most important rules across five language editions were created by a relatively small group of users who claimed their decisions reflected community consensus. In these cases, users developed templates that adopted the technical authority of documentation to translate their individual judgments into seemingly collective ones. To understand this process, it is necessary to consider how these practices of rule development are infused with the mechanisms of asynchronous editing and revision control as well as Wikipedia’s unique visions for collaboration and authority.
Wikipedia’s policy environment
Within popular online spaces, perhaps the most common form of governance that users encounter is characterized by centralized platforms such as Facebook, Twitter (X), and Google (Van Dijck et al., 2018). An alternative form exists in the self-governance and collaboration found within commons-based peer production projects (O’Neil, 2009; Reagle, 2010). Among the many projects dedicated to this form of governance, Wikipedia is perhaps the most popularly known. In many ways, Wikipedia’s unique form of governance begins with its sociotechnical affordances. To start, the wiki software encourages users to edit nearly every page of the project; it requests that users discuss significant changes and decisions on each article’s ‘talk page’; it demands that each edit is released under a public license; and it requires that these edits are stored on each article’s ‘history page’— a public chronicle of each edit and its author. With these mechanisms and the shared goal of producing a free encyclopedia, Wikipedia’s participatory affordances have crystalized as shared practices of transparently ‘[t]inkering with productive freedom’ and reworking digital materials so that they are legally and technically freed from their copyright regimes (Pentzold, 2021: 826, 824).
The institutionalization of these shared practices started out as proposals on discussion boards, drafted on user pages, or essays that had little or no influence on the broader community. But over the first decade, users frequently referenced these pages during moments of disagreement and misunderstanding. To indicate the growing significance of these documents to resolve repetitive concerns, Wikipedians developed attention-getting notes commonly known as “hatnote” templates and placed them at the top of their rules (Wikipedia, 2022a). These notes are ‘a piece of wiki code that creates a visual marker—often a text box with a different background color’ and ‘often alert Wikipedia readers and contributors to the current status’ of a page (Viégas et al., 2007: 450). For rule pages, these hatnotes quickly communicate to users whether the document is a policy, a guideline, or some other supportive document (Figure 1). By using this networked system of documents, templates, and hyperlinks, users created Wikipedia’s ‘policy environment’ (Kriplean et al., 2007: 2).
An example of an English-language edition rule page’s (1) title, (2) its access level, and (3) status hatnote.
Together, the policy environment of documents is the material basis for how English-language Wikipedia (2009b) certifies its claim on autonomy which is enshrined in the two sentences that have not changed since 2009: ‘Wikipedia is a self-governing project run by its community. Its policies and guidelines are intended to reflect the consensus of the community’. This perspective resonates across Wikipedia. From our own translations, the Spanish version confirms that policies ‘are intended to reflect the consensus of the editing community’ (Wikipedia, 2023a); the content of the Dutch-language policies ‘enjoys great consensus among a large group of users’ (Wikipedia, 2023b); and the French and Arabic-language hatnote templates for rules and policies state that any changes to a rule ‘need to show consensus’ (Wikipedia, 2022b) or are ‘adopted after discussion and consensus around it’ (Wikipedia, 2023c). Clearly, Wikipedia’s rules are imagined as standing in for community consensus and not the judgment of a single user.
To put it another way, like Wikipedian articles, Wikipedia’s rules engender a form of ‘community ownership’ which ‘denotes community membership and responsiveness to social norms surrounding mutual development and collaboration on a textual product’ (Yim et al., 2024: 425). But when it comes to the development of rules (and not just encyclopedic articles), each is understood as reflecting consensus, but this relies on consensus being achieved through the adherence to the rules. This circular logic draws attention to the fact that there is a significant ambiguity about what it means to reflect community consensus and therefore raises the question: Which forms of authority were used to formalize the community consensus that these documents reflect?
Before addressing this question, it is necessary to consider the various meanings typically associated with Wikipedian consensus. What counts as community consensus is sometimes sacrosanct, such as Wikipedia’s ‘five pillars’ which are a set of policies that outline the ‘principles that inform and undergird the prevailing epistemic and social norms and practices for Wikipedia participation and contributions’ (Menking and Rosenberg, 2020: 1). They announce that Wikipedia is an online encyclopedia; written from a neutral point of view; uses content that is free; by users who honor civility; and who make rational and autonomous decisions about when and how to use rules. What is remarkable is that despite Wikipedians’ capacity to make edits to any page, rule pages tend to change very little. But further, in 2008, it was recognized ‘that policy editing is slowing, and the process for adding to the policy environment is becoming more formalized’ (Beschastnikh et al., 2008: 29). More recently, Sohyeon Hwang and Aaron Shaw updated Beschastnikh’s initial findings in their study of 780 rule pages from English, French, German, Japanese, and Spanish-language editions. They found that during these edition’s lifetimes, ‘rule-related revisions become less frequent, get smaller, and tend to shift from rule pages themselves to talk pages’ (Hwang and Shaw, 2022: 352).
Because of the operational importance of these pages, this situation of fewer edits over time indicates that these pages are incredibly stable documents. And they need to be. Without the agreement that these policies are foundational and that all changes to them need careful consideration, Wikipedia’s continued existence would be unlikely. As well, the semi-anonymous nature of contributions means that ‘interpersonal trust is replaced by trust in procedures [. . .,] and the internal order of the collaboration is partly sustained by the trust endowed in policies and their execution’ (Jemielniak, 2016: 365). This form of governance has sparked concerns about the concentration of power, especially in relation to social, technical, and cultural politics (O’Neil, 2009: 38).
For example, a community the size of Wikipedia has ‘hundreds of formal, semi-formal, and informal organizations’ that not only ‘gather Wikipedians to perform various tasks’ but may also ‘vie for control of Wikipedia’s governance’ (Konieczny, 2009: 167–168). As Konieczny claimed, the majority of Wikipedians constitute an ‘idealistic anti-oligarchic majority’ (181) and the social value of transparency and communicative capacities of the wiki software have limited some power plays that would otherwise emerge from these groups. Despite this perspective, oligarchic activities have taken root within Wikipedia’s organization; Konieczny (2017: 771–772) observed that despite apparent commitments to an ‘egalitarian ethos’ by Wikipedian collegiate court – the Arbitration Committee – it exists as a ‘small group of the most active arbitrators who seem to dominate most of the decision-making process’ and ‘do little to prevent such oligarchies’ as a result of the time constraints of voluntary work.
A second institutional concern is that Wikipedia is organized according to the economic, operational, and legal controls administered by its parent foundation, the Wikimedia Foundation. For Rijshouwer et al. (2023: 1298), Wikipedia’s bureaucratic forms are separate from the alternating processes of centralizing and diffusing power that exist between the community and its parent organization. While many Wikipedians abhor any kind of external authority, the foundation has made continual attempts ‘to increase democratic accountability’ through ongoing interactions between all involved stakeholders (1298).
Another important way that power is conferred within Wikipedia is through bots. Riedl and Halfaker (2012: 80–81) described how Wikipedians have created bots as ‘force multipliers’ that deal with the inhuman task of assessing and processing each new contribution, especially in the cases of identifying and blocking vandals. While bots exist as software translations of human decisions and policies that are ‘fully automated software agents’ (Geiger, 2014: 23), the power they inscribe is not neutral. For example, the act of automating certain processes is subject to the decisions of the bot programmer and a relatively small set of dedicated users, the Bots Approval Group. So, when bots are used to limit problematic information like vandalism, systemic biases, and conflicts of interest, how these phenomena are imagined and programmatically quantified leads to interactions that are neither categorically good nor bad but are ‘messy and multiple’ (Jiang and Vetter, 2020: 84).
Such messes are tied to the conflicts and absences that emerge from an encyclopedia that is intended to be the storehouse of humanity’s generalized knowledge. As reported by Wikimedia itself, there are numerous studies that observe that the esoteric jargon, technical skill set, geographic constraints and complex set of rules created ‘gaps’ in knowledge that make Wikipedia intimidating to new users and limit involvement (Redi et al., 2020). Furthermore, many of these rules privilege masculine and Western ways of thinking about knowledge (Menking and Rosenberg, 2020; Yim et al., 2024). Thus, even when new users from marginalized communities attempt to participate in policy reforms that would make Wikipedia more inclusive, they have often been met with resistance, as is the case when evidence from oral cultures is excluded (Avieson, 2022: 402–403) or when users have sought to address Wikipedia’s gender gap (Tripodi, 2023; Jankowski, 2024). Such conditions concentrate governing power – not in the form of an oligarchy, but through racist and patriarchal distinctions.
Power therefore exists in a number of parallel expressions: between the alternations between centralization and decentralization within Wikipedia’s ‘self-organizing bureaucratization’ (Rijshouwer et al., 2023); the role that ‘bespoke code’ (Geiger, 2014) enacts this organization; and the social dimensions that delimit who can become a Wikipedian and define the culture of the community (Jankowski, 2024). Unfortunately, these bureaucratic mechanisms have naturalized certain systemic biases of an emerging and homogeneous community. As a result, they became a direct barrier to the self-organization of the community once it had outgrown its original size. In other words, a governance mechanism that facilitates democracy at one social scale, runs the risk of becoming undemocratic if it is retained for a larger group that demands greater commitments to inclusivity, and therefore different techniques to ensure its sovereign authority and ability to self-organize.
This point is captured by the fact that Wikipedia’s five pillars and core policies gained their authority when Wikipedia was much smaller, more homogeneous, and might be described, as Joseph Reagle (2010) did, as a ‘technical community’ aligned with ‘geek culture’ (47, 70). By the end of 2005, when many of the core policies were present on English-language Wikipedia (Beschastnikh et al., 2008: 31), the number of registered users was approximately 600,000—not even including the many anonymous accounts (Wikipedia, 2005). Almost exactly 18 years later, this number had risen to 46.5 million (Wikipedia, 2023d).
This situation of Wikipedia’s scale and homogeneity was famously brought to attention through a survey of Wikipedians that reported that 87% of editors were men (Glott et al., 2010). Following the report, studies were conducted on the core policies concerning neutrality and civility and found that they were used to legitimize a culture that is both ‘“resistant to female participation” and “hostile to information that challenges forms of male privilege”’ (Lam et al., 2011: 9; Peake, 2015; cited by Menking and Rosenberg, 2020: 6). Likewise, the striving to create policies that establish a coherent Wikipedian community ‘implies inclusion’ but ‘it can often entail exclusion’ (Tamani et al., 2019: 4). This point has been leveled at Wikipedia’s consensus-based form of self-governance which ‘orients the platform towards coercive forms of hegemony under the guise of “community”’ (Jankowski, 2022: 13). These kinds of exclusions and coercions have been observed with policies on undue weight (Peake, 2015), reliable sources (Berson et al., 2021), deleting articles and notability (Tripodi, 2023), as well as the guidelines concerning being bold and assuming good faith (McDowell and Vetter, 2022: 83). The Wikimedia Foundation itself has also acknowledged these concerns when a former executive director of the Wikimedia Foundation expressed when they stated that ‘[e]veryone brings their crumb of information to the table [. . ..] If they are not at the table, we don’t benefit from their crumb’ (Cohen, 2011). In a 2020 report, they identified that the policies that govern the site contribute to ‘knowledge gaps’ and exist as ‘major barriers for diverse content inclusion’ (Redi et al., 2020: 25).
While many solutions have been put forward, the one that took precedence was the task of increasing the diversity of Wikipedians. Yet, diversity advocacy often ignores the ways that technological spaces are implicated in the issues of diversity. While it is ‘important to have bodies at the table,’ Sarah Banet-Weiser (2018) argued, ‘their mere presence doesn’t necessarily challenge the structure that supports, and builds, the table in the first place’ (12). So, approaches to diversity advocacy on Wikipedia need to move beyond diverse inclusion or “expressive individualism” (Fian, 2024: 14) and must address the question of ‘how technology itself is implicated in projects of social sorting and domination’ (Dunbar-Hester, 2019: 22, emphasis original).
This raises important questions. At which critical moments in the design of policies did their authority become embedded? Which kinds of authority are enabled by these designs, and how do they relate to concerns about the inclusivity assumed in notions of community consensus? Asking these questions mean addressing Wikipedia’s policies, not as neutral texts that simply record the results of deliberations about community norms and practices, but as an infrastructure of documented authority. To do so, we observed how self-governance exists as a sociotechnical infrastructure that mediates existing asymmetries of power (Ford and Wajcman, 2017), and that manifests in the cultural technique of the document (Gitelman, 2014; Vismann, 2008, 2013). Furthermore, the political actions enabled by the policy environment are shaped by its affordances (Davis, 2020; Nagy and Neff, 2015) to provide the material basis for different kinds of authority (O’Neil, 2009).
If policies are intended to reflect community consensus, to what degree can this reflection be traced to critical discourse moments when templates were used to communicate their status as a policy, guideline, or other designations? Within this inventory and development of techniques, do they allow for democratic forms of authority – ones that would encourage inclusion – or do they enable individualistic forms of authority that complicate the imagined affordance of policies reflecting community consensus?
In answering these questions, we argue that while policies are imagined to have the affordance of reflecting the sovereign authority of community consensus, Wikipedians have not developed techniques to do this kind of political work at the current scale of Wikipedia. And so, clarity about which tools Wikipedians use to make collective decisions will provide more granular details about the politics this community has cultivated. Through a combined discourse analysis and qualitative content analysis of the history of edits made to 15 community-significant policies across five language editions (representing five of the 15 largest editions: English and Spanish, as well as Arabic, French, and Dutch), we find that the techniques used to designate the authority of a policy are most often supported by editorial document techniques. Across each of these policies, Wikipedians used the techniques for voting, polling/surveying and deliberating. However, these techniques were not used at a scale that renders the entire community legible in the decision-making process. Instead, a very small number of Wikipedians used meritocratic techniques to determine consensus, rather than reflect it. As such, we conclude that if the democratic value of inclusion is to be valued on Wikipedia, it must be integrated by way of large-scale democratic techniques that can be used to make the community legible in the policy environment.
Theoretical framework
Conceptualizing Wikipedia’s policy environment from the standpoint of media studies and science and technology studies requires some theoretical considerations of the politics associated with infrastructures, cultural techniques, and affordances. For Heather Ford and Judy Wajcman, Wikipedia’s software systems function as infrastructures for knowledge that ‘produce power relations’ (Ford and Wajcman, 2017: 7). They argued that the design, maintenance, and discussion of the infrastructure ‘requires particular types of skills and expertise’ and ‘particular types of knowing,’ each being activities that have been historically constituted as power relations that benefit men (7).
A contrasting approach comes from a tradition within media studies that argues that the power emanates from the tools themselves. One of the most compelling descriptions of this approach comes from the legal and media scholar Cornelia Vismann (2013). She argued that the making and meaning of law is not purely determined by the sovereign people involved in the process. ‘The place of assembly, the object of disputation, and the rules of decision,’ she explained, ‘also play a part in the production of the actual law’ (85). But she pushed this thesis further to argue that to exercise any kind of power, the law must first be mediated through a symbolic, social, or material tool. Therefore, what counts as the authority of the sovereign subject will always need to be traced back to ‘questions about how things and media operate’ (88). In fact, the sovereign subject must share the throne with things and media, or be dethroned by them (86).
In her historical work on files as law media, Vismann provided a case study of documents as cultural techniques. For her, these media ‘are made to impress. Their letters are signs of power; their very appearance represents the authority of the issuer. The layout of a document is a “gesture of power”’ (Vismann, 2008: 72). Furthermore, documents imbue longevity, for ‘[w]hat they proclaim counts for all ages’ (71). They are also accompanied by various kinds of seals and signatures, both authenticating their veracity as well as ‘attesting to the binding force of documented law’ (73). In this way, they become, as Lisa Gitelman (2014) described, ‘epistemic objects’ that have a unique recursive function: they enable ‘the kind of knowing that is all wrapped up with showing, and showing wrapped with knowing’ (1). They become self-evident evidence, a certifying form of authority. Given that Wikipedia’s policy environment is dependent on a database of files and communicated through documents, Vismann and Gitelman’s theories are particularly useful for analyzing the community’s tools for self-governance.
Bridging the gap between the power relations of technically adept agents and the effects of cultural techniques is Jenny Davis’ (2020) ‘mechanisms and conditions framework’ for analyzing affordances. For her, the mechanism side of affordances identifies a set of deontological actions: ‘technologies request, demand, encourage, discourage, refuse, and allow particular lines of action and social dynamics. Requests and demands are initiated by the object, and encouragement, discouragement, and refusal are responses to subjects’ inclinations’ (Davis, 2020: 11). However, these mechanisms need to be incorporated into personal and social settings to become an affordance. Therefore, making sense of a mechanism can ‘vary by perception, dexterity, and cultural and institutional legitimacy’ (11). One can add to the cultural component of the framework the fact that some affordances are imagined, which ‘emerge between users’ perceptions, attitudes, and expectations; between the materiality and functionality of technologies; and between the intentions and perceptions of designers’ (Nagy and Neff, 2015: 5). This leads to situations where both users and designers’ ‘affective experiences’ are shaped by the effects and agencies a technology is expected to produce (2).
Framed through these approaches to infrastructure, cultural techniques, and affordances, Wikipedia’s policies perfectly align with what Mathieu O’Neil (2009) defined as the online-equivalent of democratic sovereign authority, a type of authority where ‘people can justify their actions by referring to a formal social contract,’ and that thus ‘derives its legitimacy from the general will’ (74). Under this form of authority, sovereignty is given the institutional ‘material basis [. . .] which can make or interpret predictable and enforceable rules about the propriety or impropriety of certain actions’ (emphasis, original: 75). However, with ‘online tribal bureaucracies’ like Wikipedia, authority is also derived from a merit-based authority that stems from a ‘group’s affective recognition of outstanding work’ of specific individuals (37) and can be defined as ‘mastery of a frame’ (Tkacz, 2014: 86). Originally, this took the form of ‘hacker-based charisma,’ but since many hackers also maintained Internet systems, their character runs parallel with both administrative authority and editorial authority (O’Neil, 2011: 2; 2009: 153).
Individual freedom and merit were not only popular social principles for many online communities, but Wikipedia’s adherence to them can be traced back to the Objectivist philosopher Ayn Rand and the neoliberal economist Friedrich Hayek. These two philosophers directly influenced Wikipedia’s co-founders Jimmy Wales and Larry Sanger and were cited as inspiration for the social design of Wikipedia (Reagle, 2010: 57; Mangu-Ward, 2007). As a result, Guilherme Fians identified that Wikipedia’s founders attempted to solve the neoliberal problem of ‘minimizing top-down hierarchies’ through these ideas to maximize ‘the autonomy and power of each individual over themselves – and themselves only’ through the platform. In doing so, the process of ‘becoming Wikipedian’ encourages embracing ‘technoliberal understandings of authority and authorships’ (Jankowski, 2024: 17, 4). Indeed, the collective authority of the community is a complex practice that involves an amalgamation of Deweyian, Habermasian, and Hayekian notions of consensus that are at often at odds with feminist theories of democratic dissensus (Jankowski, 2022). In a practical sense, this means that ‘Wikipedians perform consensus: not only through understanding and decision-making, but also through acts of composing, showing, processing, closing, and calculating’ that do not make room for difference (1). With these concepts of authority in mind, we interpret Wikipedian claims of ‘community consensus’ as an ‘ideal technique’ that is assumed to exhibit democratic sovereign authority, and we analyze the degree to which this claim holds up to scrutiny.
Methods: discourse analysis, qualitative content analysis, and digital methods
In alignment with our questions about community and authority, we relied on discourse theory to analyze how identity and group formation are articulated through meaning (Jørgensen and Phillips, 2002: 44; 50). We mobilized this theory through a qualitative content analysis using the four iterative and reversible steps of sampling, collecting, analyzing and interpretating data (Caliandro and Gandini, 2017). In addition, the ‘content’ that can be analyzed from Wikipedia is unique because discourses on group formation are not limited to the production of words alone but also include Wikipedia’s ‘bespoke code’ (Geiger, 2014: 4): templates, images, and metadata that are networked across project pages through talk pages and history pages. Based on these conditions, we followed the advice of Caliandro and Gandini (2017: 194) to be sensitive to the way that social media shapes content production which includes being cognizant of how metadata can be used in this context of filtering and sorting data.
Corpus, sampling, and collection
Like Hwang and Shaw (2022), we also examined Wikipedia’s multilingual policy environment. However, while they conducted a content analysis to answer quantitative questions, our questions about meaning and group formation could not be automated in terms of collection and translation. As such, we created a purposive sample of rules from across all language editions. Again, simililar to Hwang and Shaw (2022: 349), we focused on Wikipedia’s largest language editions and then selected from the top 15 largest language editions based on the researchers’ native fluency in these languages which included the Arabic, Dutch, English, French, and Spanish-language editions. Because these editions exist as some of the largest language editions of Wikipedia, they serve to represent many of Wikipedia’s most prevalent practices of rule development. However, they cannot be said to represent all of Wikipedia, a position that is often attributed to the English language edition.
We began by ranking all policies and guidelines for each language according to five equally weighted dimensions: visit frequency; wikilink frequency; a start date during or after 2005 when personal interactions were ‘no longer sufficient for making decisions,’ (Reagle, 2019: 6); the condition that first versions were drafts and not stub pages; and whether hatnotes were used on the page. These criteria were used to capture the diversity of rule-authorizing practices while also constraining the corpus enough so that the analysis of the data could directly answer our research questions. From this ranked list, we chose the top three rule pages from each language edition (Table 1).
Purposive sample of the top three ranked rules in each language edition.
Data collection began November 2022 and ended in April 2023 by scraping the history page of each rule using the Wikipedia Edit Scraper which was developed at the University of Amsterdam (Digital Methods Initiative, 2016). After cleaning the .tsv files of extraneous data, we gathered a total of 8040 edits for the period of February 2005 to April 2023.
First-cycle and second-cycle coding
To answer our research questions, we combed through the corpus looking for ‘critical discourse moments,’ which are defined as moments that ‘challenge the “established” discursive positions’ (Carvalho, 2008: 166). The unit of analysis was therefore discursive statements (text, graphics, and code) that were used to designate or contest the authority of the rule page in the process of ‘reflecting’ community consensus.
After the initial step of pre-coding the attributes, we developed iterative coding schemes for holistic coding, descriptive coding, value coding, and versus coding (Reynolds, 2019: 59–60). Making these schemes coherent and exhaustive required comparing wikitext and the display of each version using Wikipedia’s ‘diff’ comparison feature, checking edit summaries and observing metadata on the history page, and reading discussion threads that were referenced in edit summaries. Through the process of first-cycle coding, we coded 147 critical discursive moments using the final coding schemes (Tables 2–5). As part of the data analysis phase, the categories of these schemes were synthesized together by creating ‘larger umbrella codes, which represent the dominant discursive formations’ (Reynolds, 2019: 61). These were developed during the second-cycle coding phase which includes a longitudinal analysis of the discourse over time and overarching patterns (61).
Coding scheme for the designations found in each rule page.
Coding scheme for the holistic coding of actions used to designate the authority of the rule.
Coding scheme for contests over the designation of a rule.
Coding scheme for types of power used during the designation of a rule’s authority.
Coding schemes
For the analysis of contestation, we followed Iñiguez et al.’s (2014: 12) definition of serious ‘mutual reverts’ which occur when two experienced editors revert one another’s edits, often described as an ‘edit war’ (Wikipedia, 2003). Using these reverts as a starting point, we defined the end of the contest as two weeks after the last designation edit. The character of these edits was then identified by examining the edit summary, content, references to talk pages, and vote pages. In the case of talk pages, we searched for keywords used in the edit summary or the specific edit to find associated discussion threads and associated users. The final coding scheme draws on concepts from the theoretical framework to deductively assess which types of authority were being invoked to legitimate the act of designation.
Limitations
While our study covered some of the largest language editions on Wikipedia, it was limited by the sample of rules in these specific languages and our ability to translate and interpret them. Furthermore, our research relied exclusively on Wikipedia’s publicly available data. At issue is the fact that some data may have been deleted, conversations may have occurred ‘off-wiki,’ and users do not write everything they think. As such, we encourage future research to ask questions about rule development that involves other language editions, interviews or surveys with Wikipedians about power imbalances in rule making, as well as compare these processes with other commons-based peer production projects and platforms.
Data analysis
Longitudinal analysis
There are two points of interest to note about the timeline (Figure 2). Considering the length these rules have been in place, the period of developing their authority is relatively short with about half of the rules being settled within a year of the first edit. The remaining rules take a variety of years to shift, often with no contestation. Two exceptions come from Dutch-language Wikipedia where the rule ‘Balance’ was never designated, and ‘Be Clear and Concrete’ remained at the proposal stage. The second observation is that by 2011, the designation of 11 of the 15 had been settled.
Timeline of the period of developing the authority of rules.
These findings support and augment previous studies. While Berson et al. (2021) reported that the largest number of edits on the English language ‘Reliable Sources’ occurred between 2006 and 2009, it was during this period that the rule was designated as a guideline. So, while the content of the rule underwent its ‘[g]reatest editorial changes between 2011–2020’ (Berson et al., 2021: 14), its status was already established in the first few months. Likewise, our findings confirm Hwang and Shaw’s (2022: 351) observation that rule creation happens largely within ‘years 2–6’ of the lifespan of a language edition. However, the authority of the rules we studied are often established in a much shorter timeframe, from just a few months to a year. In other words, the authority of a rule is often determined long before the actual substance of the rule is agreed upon.
An analysis of the type of edits used to designate authority presents two approaches to rule development. The first approach is one that has very little contestation which is common for the Dutch, French and Arabic-language rules (Figure 3). However, the Spanish and English-language rules are comparably much more contentious. This is especially true of the English-language rules where the technique of the revert is heavily used. This clearly identifies that the decision about the authority of English-language policies was not a linear progression from less to more authority.
Comparisons of editor actions that designate authority from the first edit across language editions.
Pattern one: rules that are primarily designated through document authority are largely uncontested, include minimal steps, and involve only a handful of editors
For most of the rules analyzed, each begins with the editorial authority of creating a page. This is followed by users deciding to denote the status of the rule by invoking the document authority of a template a few times. For the nine Arabic, Dutch, and French-language rules, there were only one to six critical discursive moments. Moments of contestations of different kinds of authority were also rare. For example, in the Arabic-language ‘Article Organization,’ there was a three-step contestation of a designation involving a revert. In addition to these nine rules, the Spanish-language rule of ‘Conflict of Interest’ also had one moment out of its total of seven when the authority of an administrator was invoked to protect the page. Combined, the authority of these rules reflects a total of 35 critical discursive moments which were produced by a total of 32 users. These low levels of participation and contestation appear to be the norm in the development of our sample of rules.
Pattern two: rules where authority is contested tend to circulate within several regimes of authority and enlist techniques associated with sovereign democratic authority, but only at relatively small scales
Wikipedians who contribute to the Spanish-language edition often use the democratic technique of voting to designate the authority of rules (Figures 4 and 5). During the designation process of the Spanish-language rule on ‘Deletion Queries,’ there were six critical discourse moments when the rule was contested. In this process, a sequence of six separate users edited, substituted, and customized templates to move it from a proposed policy to an official policy. What is unique about the final edit in this sequence is that it references a vote to decide the designation as official. On the Spanish-language Wikipedia, there is a page dedicated to archiving all voting sessions. For the ‘Deletion Queries’ decision, users had 14 days to cast their vote, and 69 users voted on the issue, with two of those voters having been involved in edits to designate the rule.
Common frequencies of authority found in 10 out of the 15 rules.
The consensus for the final designation of ‘Consultas de Borrado’ (Deletion Queries) which included five individual users sequentially processing the designation of the rule and then ending with a final decision made by 71 voters.
The contestations about the Spanish-language rule of ‘Reliable Sources’ (Figure 5) were more complex. In terms of editorial and document authority, four users were involved in a four-step cycle of edits of reversions, deletions, and replacements. However, these actions were often legitimized by limited discussions of four to seven users. After several rounds of editing and discussion, the final decision to designate the authority of the page was voted on by 38 users, which included three designating editors.
The rules for English-language Wikipedia tended to rely more heavily on discussion than any of the other languages. This was seen in ‘Disruptive Editing’ (Figure 6) which involved a small group of five editors who relied on editorial, document, and democratic forms of authority to legitimize their edits. The more extensive example of this kind of activity was observed on ‘Proposed Deletion’ (Figure 7) where four editors were engaged in contests that connected to a larger group of 29 users involved in four separate discussions. In these cases, there was much greater concentration on the link between designating a rule and the discussions involved to legitimize that decision.
The consensus for the final designation of ‘Reliable Sources’ (Fuentes Fiables) included 12 closely related steps where users discussed and templated the designation of the rule before ending within a decision made by 38 voters.
Disruptive editing’s small network of contestation.
Of the 15 rules analyzed, the English-language ‘Reliable Sources’ was the most complex. While the period between the first and second designation edit to this rule was 417 days, the next nine designation edits happened over the course of two days. Just a few days prior to this activity, two threads discussed why the guideline was not a policy and how to achieve consensus about what the guideline should do. During the two-day period, five more discussions were used to discuss these questions, with a total of 40 individuals making comments. As a result, the legitimacy of the designation was a function of talk page deliberations. Sixteen designation edits and half a month later, a new discussion assessed whether or not Reliable Sources was a guideline. This turned into a survey of 40 participants. Unlike Spanish-language rules, the survey did not finalize a decision, and the question of authority remained open.
Overlapping with this period was a set of contests over how to actually designate the authority of the rule. One group continuously used formatted wikitext to represent the authority where another group used the {{guideline}} hatnote. In several instances, the result was visually the same, but it was important to these Wikipedians to either convey the contingent nature of authority (through formatted wikitext) or to reinforce its stability by using the template code (Figure 8, Moments 19–41).
English-language ‘Proposed Deletion’ with its tight knit group of contesting editors and their supporting discussions.
The following five months included nine more attempts to challenge and solidify the status of the rule. Part of the complication was that ‘Reliable Sources’ was being prepared to merge with a rule called ‘Attribution.’ At one point, Wikipedia’s founder Jimmy Wales became aware of the situation and changed the guideline template himself (Figure 8, Moments 36–37). O’Neil (2011) explained that in this specific case, ‘since these actions were performed by the project’s charismatic co-founder, they were not perceived as unjustified. However, they contradicted the procedural basis of a sovereign authority regime and generated controversy’ (p. 282). So, while English-language ‘Reliable Sources’ demonstrated techniques of deliberation and surveying which allude to collective decision-making, the crisis ended shortly after Jimmy Wales used his hacker-charismatic authority as Wikipedia’s ‘benevolent dictator’ (282) to shape the policy environment.
Data interpretation
We observed that 10 of the 15 rules had a combined total of only 35 critical discursive moments and only 32 users. In these cases, editorial and document authority were nearly the exclusive form of authority being invoked (Figure 9). This means that the techniques used to agree on the authority of these rules were often based on three individuals who used their technoliberal authority to make an edit or add a template as the only form of legitimacy for their decisions. In this way, these actions speak to the fact that the authority of the rules was determined by an incredibly small set of individuals and did not include evidence of concrete, traceable, and conscious acts of consensus making at the community level.
English-language ‘Reliable Sources’ demonstrated the most complex system of contests between types of authority.
There were several exceptions where the designation of the authority of a template was much more involved – such as with the Spanish and English-language ‘Reliable Sources’ (Figures 5 and 8), and the English-language rule of ‘Proposed Deletion’ (Figure 7). Here, there was a greater opportunity for democratic forms of authority to legitimize the decisions, but the respective participants were 87, 51, and 33. These numbers bring into question the degree to which the authority of these rules reflect community consensus and not just heavy users. Such a conclusion aligns with previous research that observed that 77% of Wikipedia’s articles are written by 1% of editors (Matei and Britt, 2017). But we also explain this situation further.
During the pre-formalization of the policy environment, rule pages were intended to be treated like any other page on Wikipedia – they were rules-as-wikipages. In the terminology of affordances that Davis (2020) describes, users were both socially and technically allowed and encouraged to edit these pages. However, the debates over the hatnote templates were precisely the moment these pages transformed from being ad hoc wikipages into institutional documents that certified the legitimacy of Wikipedian practices. This transformation changed how these pages could be interpreted. As Lisa Gitelman (2014) argued, documents make claims about themselves that are treated as evidence of those claims. And so, these pages were interpreted by new users as self-evident and authoritative – the rule-as-document demanded to be followed and discouraged editing because they were often semi-restricted and required ‘community consensus.’ This context of rules being interpreted as documents and not wikipages helps to explain why changes in rules are so infrequent. The hatnote templates announce that users do not have the sociotechnical capacity to change the authority of rules as individual editors, while this is exactly how rule authority was first established.
Despite claims that policies reflect community consensus, these rules were constructed by a handful of users who determined what was consensus – sometimes using deliberation or voting, but most often by just making an edit. In accordance with Vismann (2013), this can be explained by the fact that wikipage editing determined political agency; since editing was designed as an individual and technical activity, the first rules were constructed from ‘consensus’ through these technoliberal politics. Such a view of consensus runs counter to democratic sovereign authority generated by a ‘general will.’ In the absence of others, the authority of the few designated itself as the authority of the many.
While Wikipedia pages have mechanisms that allow for the discovery of who made which edit and when, they often do not transparently communicate the legitimacy of each rule. In this study, the clearest form of this activity as a technique was observed on the Spanish-language forums. After their vote, the template included a link to the vote and the date that decided the authority of the rule. Another intervention could be to design policies and guidelines – not as certified documents, but also as historical accounts of precedents and shifting changes to Wikipedian practices. Perhaps one of the best illustrations of this design in practice was Beschastnikh et al. (2008) visual analysis of policy citations on talk pages. They found that ‘[e]nacting policy is one indication of participation in self-governance’ and that this activity is used largely by new Wikipedians who ‘tend to cite the most popular policies, revealing the efficacy of socialization mechanisms’ (Beschastnikh et al., 2008: 31). A rule page that displays data about historical and current talk page policy citations could – in material terms – actually ‘reflect’ community consensus. Importantly, such a design intervention would have the additional benefit of highlighting the emerging or declining significance of rules for defining Wikipedian practice. Because Wikipedians value active contribution, the longevity, currency, and intensity of policy enactment should be considered a starting place to create techniques that reflect community consensus beyond the determinative nature of authority attached to policy hatnotes.
Conclusion
To be clear, it would be folly to blame the limited participation of all Wikipedians in the policy environment on the individual editors involved with acts of rule designation. After all, this type of stochastic action is reinforced, valued, and valorized throughout the site’s technoliberal logic. Instead, this situation is a collective problem of misrecognizing policies and rules simply as neutral documentations of best practices.
In this article, we approached this concern by looking precisely at the way that authority was attached to rules across five different languages through the lens of cultural techniques. Due to the nature of the data collected, we focused primarily on two language-editions: English and Spanish. Since the English-language Wikipedia was the space where early rules were developed by its founders and initial users, it became the primary place to sort out the questions of what Wikipedia is, and which processes the community values. Within the English-language rules, the authority of rules was based on the editorial authority of one user, who may have found support through limited deliberation. In contrast, we found that two Spanish-language rules relied on voting as the technique to legitimize the final status of each rule. And for an additional comparison, we observed that the French, Dutch, and Arabic-language rules had very little activity dedicated to designation, with the Dutch-language edition being quite ambivalent about the use of hatnote templates. Despite this variety, most of the designations given to the rules were uncontested after the initial year of discussion.
These results support our argument that while Wikipedians rhetorically describe their policies as ‘reflections’ of community consensus, the implementation of these techniques by small groups did not have the sociotechnical capacity to enable this participatory affordance. Part of the explanation for this mismatch is that when these techniques were initially used, they were believed to be sufficient to ensure the community’s sovereign authority to self-organize. But just as Wikipedians and others began to question the site’s capacity to be radically inclusive in 2011, the rules-as-wikipages had already transformed into rules-as-documents. They were no longer adaptable to the new social conditions of an expanded community. In our investigation of this situation, we suggest that the concern of inclusion could be addressed, in part, by adopting techniques that could fulfill the imagined affordance that rules are intended to embody – to represent the community. In doing so, they would increase the legibility of the actually existing community and therefore lend greater democratic legitimacy to Wikipedia’s rules.
Footnotes
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is supported by the Wikimedia Foundation’s Research Fund under Grant No. G-RS-2204-08608. In accordance with the Wikimedia Foundation’s Open Access Policy, the data used for this research is publicly available: Jankowski S, Celis Bueno C, Kemper J and Sabbah O (2024) Wikipedia Policy Status Moments in Five Language Editions. University of Amsterdam. Available at: 
.