1The success of Wikipedia is unique in the history of the Internet. Created in January 2001, like thousands of other unpromising experiments launched all over the Web by enterprising internauts, the interactive encyclopaedia has now become an essential resource for sharing knowledge via the Internet. The tenth most consulted website, offering over a million articles in its French version, an obligatory reference for the majority of student essays, a constant source of information for internauts, a space for Internet visibility for those who seek to exist on the Web, Wikipedia has become one of the main pillars of the Internet1. Various other bottom-up innovations [Von Hippel, 2005], such as Google or Facebook, have experienced a similar trajectory, but few have succeeded in fulfilling their initial promise: refusing commercialisation, and giving each and every internaut the right to contribute to the encyclopaedia without setting up centralized a priori editorial control. If the cover provided for the various encyclopaedic areas remains very unequal [Halavais, Lackaff, 2008], comparative and statistical analyses have uncovered a disturbing paradox: with respect to the quality criteria that Wikipedia gave itself – to wit the refusal of original texts, factual verifiability, and as neutral a tone as possible – there is a strong statistical link between the quality of an article, the number of contributions and the variety of distinct editors [Wilkinson, Huberman, 2007 ; Anthony, Smith, Williamson, 2005]. Wikipedia has thus experimented on a world-wide scale a model of cooperative writing which privileges the strength of numbers over the selection of competencies [Surowiecki, 2004; Sunstein, 2006; Landemore, 2010], even if the number here is not a simple aggregate, but the precondition for putting into operation a process of mutual surveillance and coordination [Kittur & Kraut, 2008].

2In fact, Wikipedia constitutes one of the most accomplished and radical experiments in self-organized cooperation. Its functioning has never ceased to intrigue observers, and it continues to nourish the incredulous suspicions of the elite. Reminiscent of a sorcerer’s apprentice, Larry Sanger, one of Wikipedia’s two original founders, has never really understood how the whole thing was possible, and has never ceased to denounce the monster he created and which eventually escaped him. How can the reliability of the articles in an encyclopaedia be ensured if, by means of a simple click, anyone and everyone are able to erase, correct or write the articles in it? Can it really be possible that no one, no academic, no expert, no learned man or woman of reason evaluates the contributions of the masses? How could the writing of the most consulted encyclopaedia in the world have been entrusted to the ignorant? The success of Wikipedia constitutes an intriguing denial of the belief in a “natural” hierarchy of competence. As an example, it shows that it is not necessary to entrust knowledge production to institutions that organize a selection of the qualities of contributors, but that it is possible to open up a space to which all have access without condition, as long as very strict rules are set up to decentralize as strongly as possible the mutual control that participants exert on, and only on, the respect of procedures for the production of knowledge2.

HOW DO THE WIKI KEEP IN MIND THE EPISTEMIC SCOPE OF WIKIPEDIA?

3Wikipedia has become a veritable field of research with its international meetings (Wikimania), its publications, and specialized conferences (WikiSym)3. The free and open access to all the data concerning the encyclopaedia, where all the processes are public and accessible as a matter of principle, has favoured substantial scientific production in recent years. The interest of these studies lies in the opportunity of exploring a very large corpus which enables rigorous statistical objectivation of the slightest behaviour patterns, however minor, of the encyclopaedists. In the present article, we propose a synthesis of some of the results of these studies in the light of a questioning which aims to specify the sort of governance at work within Wikipedia4. The recent renewal of studies on the governance of on-line communities has been carried out with reference to the question of networks (as opposed to territorial regulation by the State, and to regulation by competition and market price mechanisms), and with reference to the question of collective management of common goods [Orstom, 1990]. These studies have made it possible to identify the institutional arrangements which favour the formation of large collectives of peer-production and exchange on Internet. They have highlighted the importance of dividing activities into modular unit tasks to enable peer-to-peer cooperation [Benkler, 2009]; the absence of hierarchical command and an a priori work-plan [Foray & Zimmerman, 2001]; the preference for consensus rather than vote for decision-making [Aguiton & Cardon, 2007a]; [Urfalino, 2007]; tolerance for varying degrees of commitment, which allows for weak as well as strong cooperation [Aguiton & Cardon, 2007b; Haythornthwaite, 2009]; the importance of symbolic recognition and reward in the motivation of the actors [Ashton, 2011]; and the need to decentralize as much as possible the power to sanction behaviour [Ostrom, 1990]. A common trait in all these studies is that they point towards the displacement of the normative source on which network institutions are based, from substantive legitimacy towards a procedural legitimacy based on processes [Habermas, 2000]. This characteristic is in line with many studies in political science which bear on the new institutional forms which have appeared in recent years including: transnational collectives and institutions [Castells, 2009], independent authorities [Rosanvallon, 2008], the regulation of the Internet [Brousseau, 2001], systems of participatory consultation [Sintomer, 2007], or new forms of political action in networks [Rumpala, 2007]. Although these self-organized collectives closely combine on-line and off-line interaction [Hess & Ostrom, 2007], they become particularly striking when they are constituted in the digital world. A detour therefore into the sociology of science and technology may be useful in order to grasp the singular impact of the proceduralisation of coordination, when the latter takes place in connexion with a digital interface.

4In a famous article concerning the way in which the cockpit, rather than the pilot, “remembers” the speed of the airplane, Edwin Hutchins [Hutchins, 1994] gave an eloquent illustration of the approach known as “distributed cognition”. He showed that it is not necessary for the mainsprings of coordinated action to rest on the human actors alone (their cognitive capacities, their representations or their interiorized schemes of action), applying “heroic” hypotheses concerning their rationality, their strategies, their motivations or their values, when the environment in which they interact (particularly complex socio-technical environments) is equipped with a whole set of artefacts (indicators, objects, signals, etc.) which guide, channel, control and check that their actions are appropriately adjusted to the aims of that environment. The cognitive activity of piloting an airplane resides not so much in the brain of the pilot, but rather in the whole system, which includes humans, artefacts and objects. The whole thrust of the analysis strategy developed by the distributed cognition paradigm [Hutchins, 1995] consists in sharing out the representations and the values which direct collective action between the individuals and the socio-technical environment within which they act. This perspective has been greatly developed in approaches of sociology of the sciences, which insist on the fact that technical devices incorporate principles, values or policies that are activated when human beings interact with them [Winner, 1986; Latour, 1994; Lessig, 2000].

5In the light of this, it is not possible to account for the astonishing stability of Wikipedia without considering the internauts together with the “wiki” that allows them to act as “wikipedians”. There are in fact many studies that ascribe strict requirements to the actors in an attempt to explicate their orientation towards cooperation. The controversial debate surrounding the encyclopaedia thus opposes two contradictory interpretations of the qualities and the motivations of the participants. For some, it is inconceivable that the wikipedians could have all the qualities of competence and impartiality required to achieve the aims of the encyclopaedia. For others, the existence of Wikipedia is precisely

6the proof that, in our societies, there is a sufficient number of persons

7who are impartial, altruistic, and motivated to produce and disseminate knowledge. United in their common presuppositions, each of these opposing interpretations fails to explain the functioning of the on-line encyclopaedia by reference to the individual qualities of the participants alone [Kuznetsov, 2006]. However, in order to explain their cooperative on-line behaviour, it is not necessary to endow the actors with altruistic, disinterested representations or with a virtuous preference for knowledge [Benkler & Nissenbaum, 2006, p. 410ff]. It is the wiki itself, with its very format, its style of writing, its discussion pages, its system of traceability of contributions and deletions, its alert messages, its tags that notify breaches of the principles of the encyclopaedia, etc., which constitutes a veritable moral interface that encompasses the necessary resources for achieving behaviour directed towards the production of knowledge. The wikipedians are not, by some stroke of luck or by themselves, born encyclopaedists. It is the wiki that turns them into encyclopaedists. This performative capacity of the interface5 is particularly apparent when one observes the trajectory of internauts in the context of the community [Bryant et al. 2005]. Their career as an encyclopaedist often begins for personal, utilitarian or contingent reasons: to correct a fault in passing, to complete an article by adding new sources, to write a missing article on a subject which concerns or interests them… It is only very progressively, by delving into the realm of interactions, of corrections, of watching over their own articles and those of others, that some actors become more actively engaged in the collective work of the on-line encyclopaedia and gain the title of “wikipedian”. The recognition and the internalisation of the values of the community are subsequent emergent effects of the on-line interactions, rather than pre-requisites for engagement [Aguiton & Cardon, 2007b].

8Thus it is through contact with the interface of the wiki, through interacting with other contributors, through being subjected to the remarks that others post on their contributions, that the participants yield, adapt and adjust their behaviour to the collective rules of the encyclopaedia. The sort of epistemic reason which guides the great majority of behaviour on Wikipedia is to be sought neither in individual persons, nor in the interface as such, but in the mutual adjustments which make it possible for people to interact via the wiki. Now, the interface of Wikipedia organizes a very tight contiguity between the space of practices and the space of the norms of practice [Reagle, 2010, p. 47ff]. This recursive dependence between the practice and the rules is at the heart of the reflexive functioning of the on-line encyclopaedia. It sets in place a type of behaviour which continually puts the actors at a distance from their contributions. The governance of Wikipedia can thus be characterized by this incredible exigency to displace, in the most systematic and radical manner possible, evaluations which bear on persons, their competence, their interests, their motivations or their characters, in favour of public debate on the epistemic validity of their contribution. In order to reflect this process, we wish to highlight five lessons to be learned from the collective experience of Wikipedia.

LESSON N° 1. [[WP: IGNORE ALL THE RULES]]6 – TRANSFORM THE FOUNDING SPIRIT INTO PUBLIC AND ACCESSIBLE RULES

9Wikipedia was not born radical. Before it, Nupedia, the first on-line encyclopaedia launched by Jimmy Wales in the context of his society, Bomis, was very traditional in form. A head editor, Larry Sanger, specially recruited for the task, overviewed and validated the contributions of authors chosen for their expertise [Levrel, 2006]. The process of editing and validating the articles was very rigorous and based on a college of certified experts. An anecdote recounts that Larry Sanger verified that the contributors did have a doctorate (PhD), requesting that they send him their degree by Fax [Lih, 2009, p. 38]. After a year, Nupedia had managed to collect no more than 20 articles: the process was drawn out, the investment of the experts only amounted to dabbling, and the financing of the project was fragile. Nupedia would have been quickly forgotten if Larry Sanger and Jimmy Wales had not decided in January 2001 to launch, as an experiment, a wiki aimed at drawing up rough first drafts of articles for Nupedia, allowing experts and others to write, correct and annotate the texts. With astounding rapidity, Wikipedia, initially a simple “play-ground” for Nupedia, developed strongly and attracted widespread attention from non-expert internauts, whereas Nupedia itself was definitively abandoned.

10The earliest stages of Wikipedia were decisive for the invention of its regulation. The definition of the five Founding Principles of the encyclopaedia7, the 38 (12 for French Wikipedia) Rules governing the writing of articles and their codification in 189 Guidelines (38 for the French version), are indeed a direct consequence of the transformation of the spirit of freedom without rules, which was shared by the first circle of experts, into a set of public procedures appropriate for receiving contributions from anyone at all. In the detailed account that he has given of the initial period of the on-line encyclopaedia, Larry Sanger [Sanger, 2005] tells how the first months of Wikipedia were feverish moments of «wiki-anarchy», a sort of adhocratie where it was not necessary to formulate explicit rules in order to find a de facto common agreement. The small group of contributors, most of them University academics who had already participated in the abortive attempt at Nupedia, shared the same meritocratic ethos. These members considered the absence of rules as one of the reasons for their engagement. Larry Sanger then formulated the famous rule: [[WP: Ignore all the rules]]8 which came to serve as a founding myth for the community, even if this declaration of independence with respect to all rules contradicts the subsequent evolution of practices on the encyclopaedia. In this respect, one can see in this founding moment the same spirit of creative freedom and tacit, unconditional confidence which presided over the meritocratic-university model of the beginning of the Internet itself [Cardon, 2010]. The quality of the contributors, their common social and cultural origins, and their proximity with the academic world made it highly unlikely that there would be any uncivil behaviour. The ethos of the contributors formed a spirit of community, and served as a rampart against its possible distortion. However, this initial state could not resist the test which came when the project became a mass undertaking. The “anarchist Eden” came to an end in August 2001, when the very success of Wikipedia, as relayed by the site Slashdot, brought in a large number of new contributors [Firer-Flaess, 2007, p. 5]. The new arrivals, with less academic qualifications, keen on technological culture, often disrespectful of the courteous exchange of arguments, threw themselves into free-for-all debates. They were sometimes aggressive, made unseemly propositions, and threw themselves readily into writing fresh articles. The historians of the Middle-Ages were quickly alarmed by the arrival on their doorstep of a horde of geeks drawing their references from the culture of gothic games and heroic fantasy [Rosenzweig, 2006]. These “barbarian invasions” contributed to the frustration and flight of the initial contributors, notably Larry Sanger, whose resentment with respect to Wikipedia dates from the controversial attacks he had to endure in incessant flamewars with new arrivals [Lih, 2009, p. 171ff]. Now, it is precisely because of this transformation in the sociological make-up of the community that the founding values were progressively formulated to become written rules which could be activated as needed.

The constitutional moment of the on-line communities

11The explicit formulation of norms of writing and the principles of governance of Wikipedia is thus to be understood as a progressive externalisation, in the form of rules and an interface, of the ethos of the original contributors. Once the principle of perfectly open publication was put into practice, they could no longer count on the behavioural familiarities of their own relation to knowledge and meritocratic values; because of this, they chose to entrust the interface itself, endowed with a few basic rules, with the task of ensuring that the behaviour of the wikipedians be indeed directed towards epistemic knowledge. The first lesson that can be drawn from the experience of Wikipedia is that of succeeding the delicate transition of a “change of scale” in Internet communities when they have to codify their guiding spirit. This process has occurred with the appearance of FAQ (Frequently asked questions) on Usenet [Kollock & Smith, 1996], or in the world of free software as for example on Debian. In the case of Wikipedia, this work of codification was undertaken progressively, with a close interweaving between the formulation of public rules of behaviour and transformations in the architecture of the wiki. In a remarkable way, the endowment of Wikipedia with a written constitution by means of a set of Principles, Rules and Recommendations, went hand-in-hand with their deep-seated integration in the conception of technical artefacts which were visible and could be activated in the very course of writing and discussing the articles. The central discussion list was dismantled in favour of a decentralized system so that each article could have its own specific discussion-page. The policy of headings and banners was set up so that the standard signals of the community could easily be put on any page. The watchlist provided a personalized tool whereby surveillance of the encyclopaedia could be distributed over all the contributors. The “three reverts” rule, which prohibits a user from obliterating the same text 3 times within a period of 24 hours, was progressively tested and put into operation in order to calm any editing wars. Robots were invented to carry out a whole series of correction and notification tasks. Tools for supervising new modifications were developed to equip the task of the RC patrol (Recent Change) which keeps watch 24 hours a day on possible vandalism, etc.

12The institutionalisation of the governance rules of Wikipedia is an ongoing process. It is, indeed, preferable that the community decides collectively and continuously on the evolution of its norms of functioning, so as not to have the impression that they have been fixed once and for all by an authority which progressively appears as being more and more external [Orstom, 1990]. The fifth pillar of the policy of Wikipedia states that with the exception of the four other pillars, all the local rules of functioning can be revised. All wikipedians can, notably in the form of Essays, put forward propositions for new rules which are then put to the vote of the community. During the developmental phase of the encyclopaedia, many such modifications occurred. Posts with specific status appeared in the encyclopaedia with the election of administrators and bureaucrats. In July 2004, the Arbitration Committee (ArbCom) made its appearance. In 2005, following the controversy concerning the biography of John Seigenthaler, a journalist to whom an article in Wikipedia attributed responsibility in the assassination of John F. Kennedy, the Recommendation bearing on the “Biography of living persons” was drawn up [Beschastnikh et al. 2008] ; [Forte & Bruckman, 2008] ; this was also the case for the procedure making it possible to promote “Featured Articles” [Viégas et al. 2007]. Thus, there have been many modifications to the Rules and Recommendations over the course of the encyclopaedia’s evolution. However, today it appears increasingly difficult to shift the central edifice of the Rules of the encyclopaedia; on the other hand, there is a decentralisation of the normative production of wikipedians towards thematic gateways for which they define local rules concerning page set-ups and editorial formats [Forte & Bruckman, ibid.].

LESSON N°2. [[WP: COMMENT YOUR MODIFICATIONS]] – PARTICIPATING ALSO INVOLVES KEEPING WATCH AND SANCTIONING

13The freedom to write on Wikipedia is inseparable from the freedom – and the duty – to apply sanctions. The majority of debates concerning the participatory encyclopaedia focus on the first freedom, and do not take into account the second. We are so used to the centralisation of the power to control and sanction that this power escapes notice when, decentralised to the level of individuals, it is exerted locally and in diffuse fashion via the practice of “revert”, a soft sanction that consists of cancelling a recent contribution by returning to the preceding state of the text via a simple click. Anybody can write on Wikipedia; but anyone can also wipe out what someone else has just written. It is this combination of rights to write with rights to keep-watch/criticize/sanction which places participants in the orbit of an argued confrontation of viewpoints. The point is that when it is collective and multi-authored, writing necessarily involves discussions and a set of operations of coordination with other authors. As indicated by Viégas et al. (ibid.), on Wikipedia, it is often preferable to “discuss before writing”. Each new arrival on the scene of the encyclopaedia undergoes a singular experience. He imagines that he has just written the definitive definition, biography or text that he is interested in; but hardly has he submitted his contribution that he finds his words corrected, revised, wiped out, and above all commented and questioned on the discussion page devoted to the article. And before he knows what is happening, a robot has formatted his footnotes, a title “this is a draft” appears on the heading of his article, and automatic inserts appear in his text asking him to reference the source of this or that assertion, or to acknowledge the “non-neutral” character of a development. Once the surprise is over, if he really wishes to participate in Wikipedia and see his contribution retained by the encyclopaedia, he will have to enter the discussion, reply to the questions which have been raised, justify his statements, accept that others touch up his text, add further references, or recompose the plan of his article. Finally he will have to watch over his text and defend it against future questionings. Numerous studies have given evidence for the statistical link between the quality, the level of pertinence and the reputation of articles on the one hand, and on the other hand the number of editorial comments on the text and the density of the discussions that have taken place [Viégas et al., 2007; Stein & Hess, 2007; Burke & Kraut, 2008].

Local, the sanction is an educational gesture

14It is the very architecture of wiki which organises and directs the way in which the writing of an article is subject to discussion. The Edition and Discussion pages, as well as the traces of each action taken on the encyclopaedia, are juxtaposed with each other. The corridors in which the modes of production of the article are discussed are close to the scene on which it will be displayed. Now, as Lave & Wenger [1990] have emphasized in their analysis of “communities of practice”, there is no better technique for learning than looking at what goes on in the workshop. This familiarisation by “peripheral attention” towards the actual site of fabrication is much more effective, and less intimidating, than being confronted with a manual of abstract, explicit rules. The wikipedians learn by watching others discuss. It is by making small contributions to the articles that they start to discuss with actors who are more experienced than they are, and it is by discussing that they come to propose more substantial contributions to the articles. Moreover, the access to the open corridors of Wikipedia is also equipped with a set of technical artefacts which make it possible to insert the rules of the encyclopaedia in each of the inscriptions made on the wiki. A discussion on Wikipedia is like a sharp dialogue between friends on subjects where they disagree. But rather than letting the arguments drift off into the delicious fuzziness which constitutes the charm of late-night free-for-alls, on Wikipedia each participant punctuates his comments by citing the article of the code which justifies his assertion or his criticism. The wikipedians continually notify each other of the errors made by citing the rule which has not been respected. An insult will surely provoke a reference to the principle [[WP: No personal attacks]] or [[WP: Respect others]], a comment on a missing source by a reference to [[WP: Quote your sources]], the criticism of a passage which is too subjective by reference to [[WP: Non-neutral point of view]], etc. Beschastnikh, Kriplean & McDonald, [Beschastnikh et al., 2008] have studied the whole set of references to the rules which are liberally scattered over the Discussion pages of Wikipedia. They show firstly that the principle of explicitly recalling the rule in question has become more generalized as the encyclopaedia has developed, to the point of becoming an increasingly required norm of “good conduct” in the community. They then emphasize, like Viégas [Viégas et al., 2007], that the more an article engages a large number of authors or deals with a sensitive subject, the more the number of references to encyclopaedia rules increases.

15The second lesson that can be drawn from the experience of Wikipedia is that individuals are more ready to participate in the production of a resource if they have a power of surveillance and sanction over the other members of the community [Ostrom, ibid.]. The effectiveness of local control is all the greater when the proximity between the culprit and the watch-keeper makes it possible to implement a set of sanctions that are so gradual that their initial manifestations correspond to a simple reminder. The culprit is not “punished”, but he uses the signals that the community addresses to him in order to learn, to revise his behaviour and to familiarize himself with the shared rules. Keeping watch and sanctioning, when they are exerted at a very low level, in a manner that is both light and public, reinforce the mutual confidence and the values of the community, much more than they serve to exclude wrong-doers. They nourish a framework of discussion in which critical vigilance is directed towards the search for an agreement supported by arguments.9 It is only when faults are systematically repeated by the same person who has ignored the previous signals addressed to him by the community that the offender will be sanctioned more severely and may have to appear before the central authorities of the community. However, most often, a local sanction that is light and immediate, combining reprimands and encouragements, is sufficiently effective; this leads to community integration and facilitates a resolution at a very low level of the vast majority of conflicts between members. What transpires from the model of self-organisation of Wikipedia is that it is not only necessary to decentralize the watch-keeping and sanctioning as strongly as possible, but that it is useful to set up a graduation of sanctions between the local and the centre, so that the correction of behaviour occurs locally (Lesson n°3) and the punishment of individuals occurs centrally (cf. Lesson n°4).

LESSON N°3. [[WP: NO PERSONAL ATTACKS]] – SEPARATING PERSONS AND CONTENTS

16In order to avoid civil war, the decentralisation of the right accorded to all to sanction others is accompanied by an absolute requirement: to separate judgements on persons from those concerning content. The ban on [[WP: personal attacks]] bears witness to the imperative obligation, in a procedural regime, to suspend all inference concerning the personal qualities, competence, intentions or interests of others – even if this ban seems to run counter to all the evidence and to be carried to absurd lengths [Black et al., 2008]. When there is a dispute between two editors, the one who lets his temper run away with him and who resorts to personal attacks runs a strong risk of being sanctioned if the conflict is brought before the Arbitration Committee – even if he is judged to have been right concerning the content of the dispute (cf. Lesson n°4). The discussions on this subject sometimes reach dizzy heights of sophistication, since the effort made to separate persons (inviolate) and their behaviour (which can be legitimately criticized) requires from the participants in the discussion a reflexive attitude which is so far from being natural as to become sometimes “heroic” :

“Make a distinction between a description of the actions of a person, and attacks against the person. There is a difference between “Your comment is a troll” and “You are a troll”. Of course it would have been preferable to say “You just provoke people”. “Your phrase is a personal attack…” is not in itself a personal attack; it describes the action, and not directly the user. In the same way, replying in an editorial comment “in reply to the accusation of bad faith made by X” is not a personal attack against the user X.»10

17The ban on judging persons is first of all a necessary condition for preserving the radical equality between participants, in particular when the cognitive and cultural heterogeneity of the participants is a pre-given element which cannot be framed. This prohibition means that no-one can make claims based on his/her statutory authority (degrees, status as an expert, etc.) or be disqualified because of his/her status (age, sex, membership of a particular community, special interest in the subject of the article). It is only the actual actions of the individuals on Wikipedia which can and must be submitted to the evaluations of others. No doubt, one of the reasons for the peculiar conversational style of Wikipedia lies in the tension that the encyclopaedia nourishes between the invitation to conduct an incessant and high-level criticism of behaviour, all the while remaining absolutely non-critical with respect to the persons who produce it.

18This tension is expressed first of all through the meritocratic logic which presides over the designation of administrators. The candidacy for a post of administrator is decided by means of a public vote which is open to all the participants for one week. In a study on the profiles of 1551 candidates for the post of administrator during the period January 2006 to October 2007, Burke & Kraut [ibid.] elucidated the characteristic profiles of those who were elected. Firstly, these candidates had accomplished twice as many editions as those who were not elected. They had also produced editions in a greater variety of domains. Finally, and very significantly, those elected had produced the most “editorial summaries”. The latter exercise consists of expressing in a single phrase the change which has been made to the text in question, this summary phrase appearing in the “History of modifications” page. This initial study has been confirmed and enriched by the research of [Leskovec et al. 2010], who studied 2794 elections between 2004 and 2008, paying particular attention to the relations between voters and those elected. They show in particular that voters always elect persons who have a number of editions greater than or equal to their own; they are much less likely to vote for someone who has fewer editions than they do. These results confirm the tendency of selective matching which, in meritocratic worlds, contributes to distribute authority on the basis of a comparison with one’s own situation, and to always place at the summit of the pyramid those who are the object of common recognition.

Preferring verifiability over truth

19Because of the multiplicity and the heterogeneity of the authors, it is impossible to know whether participants have a behaviour which is directed towards knowledge and truth, or whether their behaviour is strategic. However, the process of mutual criticism should make it possible to eliminate the negative effects of a strategic policy. The procedural character of Wikipedia comes from the fact that since it cannot base its legitimacy on a community integration of values, the whole of the cooperative edifice has to rest on the respect of common rules [Habermas, 2000]. It is indeed impossible for a community as open as this one to presuppose the existence of a prior consensus between its members sufficient to make substantial decisions concerning the encyclopaedic content.11

20All substantial discussion in such an open space can only seek the source of its validity outside itself. Therefore, since it is impossible to seek this legitimacy in an authority incorporated in the qualities, the status or the competence of the authors, the wikipedians have transferred the legitimacy of their statements to external sources which carry authority12. This is why, with the notion of a “neutral point of view” [[WP: NPOV – Neutral Point Of View]], formulated by Jimmy Wales right from the start of the project, the wikipedians have adopted a polyphonic and quasi-relativist definition of truth: when there are different versions (opinions, viewpoints, theories, interpretations), the wikipedians must present all the multiple versions that are based on reliable sources, in a balanced way that is proportional to the weight of the representation of each “version” in general opinion. It is sufficient to take into account the fact that in the absence of support for a positive and substantial agreement on the veracity of the definitions in the encyclopaedia, the regulatory principles of Wikipedia, embodied in its Policies and Recommendations, bear exclusively on the procedures for the production of knowledge. On Wikipedia, verifiability is substituted for truth.

21The third lesson of Wikipedia is that the normative ideals of procedural ethics of discussion are not unattainable idealisations if they can be equipped with an arsenal of rules and technical artefacts which make it possible to constantly verify that the participants do indeed separate persons and arguments. Citing sources, balancing different viewpoints, accepting discussion, not seeking to be original, etc., all these are procedural rules for the production and validation of contents; the wikipedians continually remind each other of these rules, so that in the end they do not have to pronounce on their substantial content. On consulting thousands of pages of discussion by the wikipedians, it is striking to note that they do not discuss the validity of a statement as such, but refer to an external source which confers legitimacy. They do not suspect the incompetence of an author, but they examine closely his capacity to accept discussion and the criticism of others. They do not denigrate the interest of some for minor authors, but they insist that they satisfy the criteria of admissibility that have been laid down by the community. The procedural aim which is observed in the discussion of wikipedians is obviously always imperfectly satisfied because this ideal horizon, formal and abstract, is very difficult to achieve in practice. Substantial judgements, challenges of individual’s status, subjective points of view or arguments of authority are liberally scattered over the Discussion pages. But in the end, it is the formal arguments which, in the event of conflict, generally prevail over substantial appreciations. In this sense, one can say that the regulatory horizon on Wikipedia is of a procedural order.

LESSON N°4. [[WP: REQUESTING DISPUTE RESOLUTION]] – CENTRALISE THE PUNISHMENTS SO AS TO MAKE THEM RARE

22The regulation of Wikipedia decentralizes as far as possible all conflicts related to contents and, by contrast, centralizes in the form of a graduated system of sanctions the settlement of conflicts related to persons. In the great majority of cases, in a systematic and principled way, the wikipedians do everything possible to ensure that the settlement of conflicts is resolved by virtue of local discussion between the users. However, in some cases the editorial battles are so acrimonious that it is impossible to avoid appealing to a higher authority. When an agreement cannot be found through discussion, the wikipedians launch procedures of alert and mediation by third parties to the conflict. Anyone can offer to act as a mediator and special status in the community, such as that of an administrator, is not required. The mediator opens a specific arena to enable the presentation of arguments by each party to the conflict, encourages the pursuit of an agreement and, when this appears impossible, invites the wikipedians to proceed to an argued vote. These mediations intervene mainly in three areas: Pages to be deleted; Disagreement concerning neutrality; Dispute resolution process.

23Finally there is a third level of regulation, where actual punishments for offenders are applied. Here however, two sorts of punishments are to be distinguished, according to whether the case involves conflicts between editors or vandalism. The first procedure is extremely rare, and compares to a court of appeal when the mediation procedures detailed above have failed. This procedure is carried out by the Arbitration Committee, a small group composed of elected wikipedians who have to instigate measures against persons who are unable to resolve their disagreements by discussion or mediation. The second case is much more frequent and can be compared to a sort of police power that the wikipedians give themselves in order to control some of the recent modifications to the encyclopaedia [Priedhorsky et al., 2007]13.

“We reject: kings, presidents and voting. We believe in: rough consensus and running code”14

24The fourth lesson to be drawn from the experience of Wikipedia resides in this desire to limit conflicts between persons as far as possible, with the consequent obligation to define centralized procedures for settling conflicts between editors when all other solutions have been exhausted. Indeed, voting on Wikipedia only exists in appearance and comes rather under the heading of consultation. If the community is called on to settle a debate about a Page to be deleted, a Conflict of neutrality, or a mediation between editors in conflict, it does indeed proceed to a vote in which each participant is called upon to justify his position by a single phrase. However, the Administrator or the Bureaucrat who will have to take the final decision is not bound by the result of the votes. This decision may even contradict the majority. This distrust with respect to the electoral procedure is common to all on-line communities of the Internet. The parameters of the electors remain vague (on Wikipedia, anyone at all can vote, which could give rise to the formation of lobbies); the number of those who choose to vote is very limited with respect to the number of potential voters; the sanctions taken after the vote (in particular the temporary banishing of an IP address from the encyclopaedia) can easily be circumvented by coming back onto the encyclopaedia under another identity (sockpuppets). The system of a public on-line vote is thus, first and foremost, interpreted as a system which enables the various parties to develop their arguments, multiply the points of view, and to kindle a consensus around the strongest arguments. These procedures make it possible to put off as long as possible the necessity for the community to punish one of its members, and to make it as rare as possible to arrive at such an extreme.

LESSON°5. [[WP: WIKILOVE]] – MAKING A PROCEDURE OF HOSPITALITY

25The encyclopaedia is constantly threatened by the menace of bureaucracy. The complexity of the rules that must be observed, the incessant risk of being accused of a fault and being subjected to a “revert”, the mumbo-jumbo of the “wikijargon” which is often indispensable for following internal debates, the feeling of exclusion which can be provoked by the strong complicity of habitual users, all this has cooled the enthusiasm of more than one new arrival on the encyclopaedia, as witnessed by Gloran:

“Like many newcomers, I got jumped on for the very first article I created. In spite of the fact that it was just about a simple monument in a small village, to document my region. […] Firstly, about inextricable and absurd considerations on the law concerning texts, law that would have to be verified at the source, then at the source of the source (and so on), and then about a silly quarrel over the name of the article, whereas my version of this name conformed to the standard norms of the French language (whatever my opponent might say with his arguments which were all the more fallacious that he was just looking to make trouble). My conclusions: on Wikipedia there reigns such a regime of terror of Copyright, of this or that licence,that the admins, at least some of them, end up by discouraging contributions as if they were going to have their house sold by auction next morning”15.

26With the growth of the encyclopaedia, criticism of a bureaucratic tendency has become increasingly severe both within and outside the community. Some critics reproach Wikipedia for being overwhelmed by an ever-increasing flux of contributions which are not directly productive. The wikipedians write fewer and fewer original articles, and spend more and more time discussing amongst themselves about the way articles should be written and the way the community should be organized [Kittur et al., 2007; Viégas et al., 2007a]. Others reproach the administrators for becoming power-hungry [O’Neil, 2009]. A whole cast of wikipedians has progressively consolidated its position by exerting a power of intimidation by the rules (wikilawyering), a tendency strengthened by the rule of selective matching which contributes to reinforcing the solidarity between those who already have a high level of publications [Kittur et al., ibid.]. Taking up criticisms published on the sites of critical observation of Wikipedia which regroup those “disappointed” by the encyclopaedia, the accusation of the existence of a “cabal”, an oligarchy which monopolizes the power, has become common currency [Canivenc, 2010].

27The bureaucratic risk is a consequence of the reliance on procedures. Since they allow themselves neither to reproach persons nor to discuss the substantial content of the articles, the wikipedians are indeed obliged to set up a very formal tone of conversation, sometimes finicky on trivial details, often demanding and harassing. This risk is not unrecognized by the wikipedians themselves. In an attempt to face up to it, they have developed a specific antidote, the wikilove, which is an invitation to respect and love which proceduralises the sense of hospitality in the form of a recommendation whose non-respect can itself be notified (!) The famous recommendation [[WP: “Don’t bite newcomers”]] fulfils this function. It enjoins wikipedians to:

“Avoid all hostility to participants who are beginners. Remain polite and respectful, and do not forget a presumption of good faith: the rules and the technical functioning of Wikipedia are only obvious to those contributors who know them already”.

28But, wikilove is not required to be applied to new arrivals only. In its deliberate naivety, it sets up a substitute ethic for the banning of making judgements on each other. In all circumstances, it is recommended not to presuppose malice in the behaviour of others, and to always presuppose that they are acting in good faith for the good of the community16. If others make mistakes, one should correct them but without assuming that they did it on purpose: “When you are in disagreement with someone, remember that it is probable that he wishes and sincerely believes that he is contributing to Wikipedia”. One should always presuppose good faith, be patient, not get angry or upset (“take a deep breath and go and take a walk”, as the encyclopaedists write in their recommendations against editorial wars), because:

“You should not act as though their errors were deliberate. Correct them, but do not judge them. Consider these persons as wikipedians through and through who are not seeking to destroy the project. Thus, even if it seems to you that they are manifestly in the wrong […], never forget that you are in disagreement with them but not at war. There will be persons with whom you will find it difficult to collaborate; that is not a reason for refusing the presence of their work on Wikipedia, neither a reason for attributing bad intentions to their actions”17.

29In this way, Wikipedia contributes to extending certain traits which are characteristic of the “hacker” ethos, to the general public. In her work on Debian, Gabriella Coleman [Coleman, 2005] insists on the tension which spans the community of free software developers between, on the one hand, the valorisation of individual autonomy and merit and, on the other hand, the values of humility, detachment, generosity and civility that are also promoted by the community. Managing to reconcile these two contrasting tendencies is vital to the success of Wikipedia. The ethos of a wikipedian places value on a sort of gentle authority, of modest recognition, of generous activism and forced hospitality. This tension is, in its way, a consequence of the type of governance which has become established in Wikipedia. The only form of legitimate hierarchy within the community must rest on the level of commitment of those involved. However, and even if they deny it, this de facto authority acquired through the strength of commitment reflects on the individuals, and gives them a specific personal status. Thus, the authority of those who are the most strongly engaged can only be fully recognized if it is accompanied by guarantees of a proximity and an openness towards new arrivals, which obliges these authorities to mask or to reduce the attributes that the community has recognized in them – an attitude which, at the summit of the hierarchy, is embodied in the figure of the “benevolent dictator” [Reagle, 2010, Chap. 6].

THE ENCYCLOPAEDIA OF THE IGNORANT

30By separating local sanctions against contents (revision) and central sanctions applied to person (banning), Wikipedia has written into its architecture an internal tension which constantly obliges the participants to remove from the discussion any evaluation of persons, of their competence and their intentions. The highly ingenious system of rules, which are both incorporated into the Wikipedia interface and carried by the vigilance exerted by the community to ensure the respect of its own procedures, makes it possible to render operational one of the most difficult cognitive virtues in the construction of an ethics of discussion. In a way, it instantiates the ideal deliberating principles by obliging persons to adopt the practice of a procedural discussion. Models of discussion on a procedural basis have often been reproached for idealising the social and cognitive conditions of their practical realization. It is indeed extremely hazardous to require ordinary persons to debate arguments while suspending all strategic or expressive activity. What the success of the system of governance of Wikipedia shows is that it is necessary to set up an architecture which separates decentralized discussion and a central power of sanction, so as to make the vigilance of the community as a whole the vector of a reasonable search for consensus.

31In The Ignorant Schoolmaster [Rancière, 1991], Jacques Rancière relates how, in 1818, Emile Jacotot was surprised to discover that his Flemish students who did not know French managed to comment, in French, the Télémaque of Fénelon after a careful study of a bilingual edition, a study guided by Jacotot who, for his part, did not speak Flemish! The students had “learnt alone and without the explanations of a master” (p. 22). The method employed by Jacotot, explains Rancière, is based on a radical “equality of intelligence” between master and pupil, which refuses a sharing of knowledge. The pupil learns alone. The master knows nothing. He does not guide. He does not explain. He contents himself with verifying the quality of the attention that the pupil pays to searching. The master is not a Socratic guide helping the pupil to find the right road within himself. The master does not know this road. He is just “vigilant”. He sets up a “relation of will” with the pupil, which is nothing like that “relation of knowledge” which always installs a “mind-destroying” asymmetry. Thus, vigilance is the only teaching of the master. “This attention which never falters” is itself what produces knowledge: “The master is he who maintains the seeker in his road, that road where he is the only one who seeks and who does so ceaselessly” (p. 58). The transformation of critical vigilance on Wikipedia into a procedure belongs to this approach. The wikipedians watch over each other.

32The community membership sets in place a “relation of will”, which invites each to watch over the searching of the others. Taken individually, the wikipedians are much less knowledgeable than the scholars. But by mutually imposing on themselves to be each the ignorant master of the others, i.e. by constantly demanding the others whether they have verified, sourced, balanced… etc. their productions, in short by checking that the others have made the requisite effort to seek, without ever questioning the knowledge of those who they press to seek, they end up bringing forth a way of producing knowledge which is more robust than that of the scholars. In a society of equals such as Wikipedia, this vigilance is not organized around a sharing of knowledge, it is radically procedural. And hence ignorant. This participatory vigilance contributes

“not to furnish the key to knowledge, but the conscience of what an intelligence can do when it considers itself as equal with all others and when it considers all other intelligence as equal to its own. Emancipation is the conscience of this equality, of this reciprocity, which alone makes it possible for intelligence to be brought forth by verification” (p. 68).

33That is surely a very apt way of characterizing this “crowd intelligence” which has caused so much debate. This intelligence does not result from an addition of items of knowledge, or from any other rule for the composition of individual knowledge, but from the collective attention that each pays to reveal his intelligence, keeping watch that all the others do the same. The transformation of discussion on Wikipedia into a set of procedures thus appears to be the indispensable condition for this incredibly audacious wager: create an encyclopaedia of ignorant people.