AUTHOR
Paul B. de Laat
ABSTRACT
Open source communities are spreading from software to other kinds of informational content. Reference works are a growing domain, with Wikipedia, the web based encyclopaedia, being the most prominent example. The OSS approach is followed closely here: everybody is invited to contribute and wiki software is used that allows collaboration by multiple parties across the Internet. Thus, an amazing feature of such communities is the amount of a priori trust that is put in total strangers emerging from cyberspace.
This particular domain of web based collaboration is an interesting field of research on the matter of trust and trustworthiness, while in practice several widely different approaches towards it have crystallized. For one thing, encyclopaedic efforts (like Citizendium, Scholarpedia and Wikipedia) exhibit remarkable differences between them in this respect. For another, within existing projects fierce discussions take place about the proper ways of handling this matter, particularly within Wikipedia.
Attention will be focussed upon ‘organizational’ means of coping with (lack of) trust: rules and regulations that channel incoming contributions. Such regulations may substitute for assuming trustworthiness on the part of participants; as a result, less trustworthiness is needed (cf. de Laat 2009). An important dimension of such ‘community design’ is the distinction between several roles: permissions granted to perform specific activities upon specific resources within the project. Let me briefly indicate some distinctions in use (‘editorial policies’). In Wikipedia, everybody without exception may directly contribute to entries that already exist (‘anonymous user’); upon registration, one may additionally add new entries (‘registered user’). In a project like Citizendium however (and equally so in Scholarpedia), policies are more restrictive: everyone may contribute (as an ‘author’), but acceptance of edits has to come from appointed ‘editors’ who oversee the development of the article involved and approve the subsequent versions. This variety in editorial policies can be described in terms of high-discretion vs. low-discretion designs (de Laat 2009).
At the same time, editorial policy is under heavy dispute within Wikipedia, both within and between the various language versions. At issue is the rise of ‘vandalism’ towards entries, and the threat this poses to Wikipedia reliability. The solution is sought in reviewing user contributions before incorporating them in the ‘stable version’ of entries. It would involve introducing more roles and layers within the encyclopaedic project, a distinct deviation from the egalitarian approach followed until now. In doing so Wikipedia would move towards a design with lower discretion, and so closer in conception to Citizendium, Scholarpedia and the like. This is not only under discussion: it has already led to a separation of ways. The German Wikipedia has adopted a review system in 2008 (with other language versions from Eastern Europe following in their footsteps), while the English Wikipedia remains vehemently against it.
Just like in OSS these design differences have to do with different conceptions of how an encyclopaedic project should be run efficiently. But at the same time, unlike the case of OSS, there is more to it: different conceptions of the role of experts and expertise also play a role. On the one hand, classical conceptions on this issue would start from a belief in knowledge as web of mutually accepted and verified meanings. In that web, experts perform a prominent role as guarantors of quality. In the set-up of a web based encyclopaedia this conception translates into certified experts as the hub for all decisions about incoming contributions. Experts adopt certain entries or fields as a whole, and steer and guard their proper development. Notice that reputation as an expert is acquired in the real world, and then imported into the virtual web based environment. Such a conception supports a design of lower discretion. Connected to this, screening for one’s credentials as an expert is common procedure for becoming a contributor. This conception, it will be argued, is dominant in Citizendium and Scholarpedia
On the other hand, more ‘egalitarian’ conceptions see knowledge as ‘distributed knowledge’: everybody is – or can be – an expert in his/her own ways. Every contribution is equal – until proven otherwise after lengthy discussions. As long as people of good intentions outnumber those of ill intentions, valid knowledge may be generated (as argued by, e.g., Cass Sunstein). Therefore an egalitarian approach is demanded from the outset. A design with a higher amount of discretion is indicated. Moreover, screening of participants is deemed unnecessary. This conception, it will be argued, is characteristic for the Wikipedia project.
Above it was observed that a ‘third way’ was developing within Wikipedia circles: a middle-discretion kind of design allowing review of contributions. It would seem that underlying these developments are not only arguments about efficiency but also about the proper role of experts. It will be argued that in a furtive fashion some conception of reputation is emerging within Wikipedia, in conjunction with systems for measuring such reputation.
Let us look at the German Wikipedia. In those circles the number of performed ‘edits’ becomes increasingly valuable. This score may translate into extra user rights. At 300 edits or more, one obtains rights to review changes made by other contributors and declare them free of vandalism. At 3000 edits or more, one obtains ‘auto-review’ rights, meaning that one’s own contributions are exempt from review. So in a crude way, by accumulating edits, some sort of reputation is acquired.
In a more sophisticated vein, software has been developed for gauging inside Wikipedia reputations. A content-driven algorithm that keeps track of the rate of change in – fragments of – text has been developed by Luca de Alfaro and co-workers (Adler and Alfaro 2007). Whenever edits performed by some contributor remain unchanged, his reputation will grow; and vice versa, when his edits are undone later, his reputation will diminish. Based on this idea – and with some more nuances added – reputations of Wikipedia contributors can be calculated on a continuous basis. It has been proposed – but not yet implemented – to use this measure for e.g. reviewing systems (only edits from contributors of low reputation needing to be flagged) and for protecting controversial articles (by barring contributors of low reputation only).
This emerging – and highly debatable – conception of expertise is noteworthy for at least 3 reasons. First, it seems to refer exclusively to the virtual world: only what happens inside Wikipedia counts. Secondly, it is ‘content-driven’: reputation is tied to whether one’s contributions remain unchanged or not. Co-Wikipedians ‘vote’, as it were, for one’s contributions by overruling them or not. This is quite different from the usual personal ratings by one’s peers. A third difference follows immediately: one acquires a reputation in general, as ‘honourable’ Wikipedian – not as concerns one (or more) specific fields of investigation. The implications of this new conception of expertise deserve a thorough investigation.
REFERENCES
Adler, B.T. and L. de Alfaro. A Content-Driven Reputation System for the Wikipedia. In WWW 2007, Proceedings of the 16th International World Wide Web Conference, ACM Press, 2007.
De Laat, Paul B. Trusting invisible strangers in open source communities: About the assumption, inference and substitution of trust. Proceedings CEPE 2009, Corfu, pp. 158-180.
Various relevant websites such as:
en.citizendium.org; knol.google.com; www.scholarpedia.org; wikipedia.org (in various languages).