Wikipedia:Trust network

From Wikipedia, the free encyclopedia
This system in a nutshell:
The trust network exists to allow users to systematically document their trust-relationships, and to see which users have declared trust in another user. It is not a popularity contest or editor rating.

This proposal for a "web of trust" is inspired by the mailing list thread that began with this posting by Jimbo Wales and was started by Pcb21 on 17 February 2004. Its implementation in the German Wikipedia as Wikipedia:Vertrauensnetz was started by Sansculotte and Elian on 23 July 2004. This proposed system for the English Wikipedia shares the three key ideas of Vertrauensnetz: giving users a formal way of declaring their confidence in other users, a way of seeing which users have declared their trust of a particular user, and the resulting structure of trust-relationships formed between all users (see below). The key differences between the inactive system and the presently proposed are that the inactive system relied upon the editing of a centrally-stored table and allowed formal expressions of distrust.

In March 2005 it was started on Meta as m:Web of trust and m:Vertrauensnetz by Arnomane.

The web of trust is neither a popularity contest nor a measurement or assessment of an editor's trustworthiness or value.[1] However, it provides an additional piece of information that may be useful when coming across another user for the first time. The Wikipedia user base is so large that two well-established and respected editors, concentrating on different areas of Wikipedia, may have no contact between each other for some time. Reading an editor's user page, browsing through their contributions, and reading the threads in their talk are valuable but time-consuming methods of getting to know someone. Discovering that several reputable users, or users that you have particular regard for, have expressed their trust in an editor is a strong indicator of that editor's value to Wikipedia. However, the sheer number of editors who trust a user should not be taken as a clear measurement of that user's trustworthiness: the fact that a user is trusted by dozens of suspected sockpuppets would only harm their reputation.

There are a variety of reasons to express trust in another user: you may have worked together on a proposal or article, reviewed many of their edits in articles on your watchlist, or know them personally. Liking another user should not generally be enough; trusting somebody requires being confident that their contributions are civil, constructive and of generally high quality. It is important that the trust network does not just become a popularity contest, and that the lack of an explicit statement of trust should never be interpreted as a statement of distrust. Additionally, it would be wise to consider carefully any thoughts of writing explicit statements of distrust, bearing in mind the no personal attacks policy.

Using the features of the trust network[edit]

To see who trusts a user[edit]

Go to Special:Whatlinkshere/User:Example user/Trusted by, where Example user is the user being considered. Alternatively, use {{trust}} with the username to produce Example user (talkcontribstruststrusted by), which provides direct links to lists of other users who trust that user, and also the contributors that user trusts.

To assert trust in other users[edit]

  1. Make a /Trust subpage of your userpage, e.g. User:Example user/Trust
  2. Include the {{Trust network}} template, writing {{Trust network|Example user}} where "Example user" is your username.
  3. After that, make a list of Wikipedians that you trust, making use of the template {{I trust|User name|Reason (optional)}} – for an example, see User:TheGrappler/Trust.

Applications[edit]

The network can help to reveal the extent to which Wikipedia contributors are trusted by their peers. This information might serve several purposes: for instance, as an indicator when considering to what extent the Wikipedia content added by a user can be trusted, when taking part in recent changes patrol, or when considering a request for adminship.

The network itself can be analysed using a trust metric to rate individual users. There are very many different ways to do this, which will produce quite different results, and it is important to note that no metric is endorsed by this proposal. The simplest trust metric is to count the number of users who trust the rated user, but this system is vulnerable to attack (for instance, the use of sockpuppet accounts to trust oneself). Another is to count how many links there are in the chain of trust between yourself and another user: if I trust A, who trusts B, who trusts C, and this is the shortest path from myself to C, then C is three links away from me. I might decide that I explicitly trust anybody one link away from me, and implicitly trust anybody up to three links away. This is very different to the previous case: the measurement is personal, not absolute, and will not be affected by sock puppetry.

More advanced metrics, most famously Advogato's innovative trust metric, are attack-resistant, but may require a core "seed" of users trusted by the metric itself. Different choices for this seeding will produce different results, and different users will decide on completely distinct seed choices: there is therefore no prospect of this producing an official editor rating system through Wikipedia using this system.

Potential for misuse[edit]

Declarations of trust should be limited to users the declarer has high confidence in, but not necessarily all those that they like or agree with. Since "who trusts you?" is more important than "how many people trust you?" there is little point in creating sock puppets to declare trust in yourself. It would not be appropriate to make statements of trust as part of deals, either for tit-for-tat recognition or for a favour elsewhere. Again, since it is unlikely that anybody offering such an arrangement is a respected member of the community, there is little benefit in having that person declare their trust in you. It is important to remember that the trust network is not a popularity contest, and so there is no need to actively seek out declarations of trust. The fact that another user has not made a declaration of trust in your favour is by no means a declaration of distrust.

A web of trust should not be overly relied upon. Naturally, new but excellent contributors will take time to be integrated into the web. Neither does the web document even widespread distrust of a user. For this reason, it will always be important to peruse user and user talk pages, as well as an editor's contributions, when trying to find out about a user you have recently encountered.

Structure of the network[edit]

Strictly speaking, the "web of trust" is actually a directed graph rather than a network. The nodes are individual users, with the relationship "User X trusts User Y" resulting in an edge directed from User X's node to User Y's. The web of trust can therefore be analysed using graph theory: a copy of the entire web of trust could be extracted from Wikipedia for this purpose.

Notes[edit]

  1. ^ The web consists simply of a collection of trust-relationships: it measures nothing. However, this information could be extracted from Wikipedia and analysed, in many different ways, to produce "trust values" for individual users. The results will vary depending on which trust metric is used, and none is officially endorsed. See the applications section for more information.