Wikipedia:Levels of consensus

Introduction
Consensus is the key mechanism for determining what content ends up in the encyclopedia, and how editors behave towards each other. It is a long held principle, both in policy and in practice, that a consensus formed by a smaller group on a topic has less weight than one formed by a larger one. These are commonly referred to as a local consensus or a global consensus but these are more relative terms than they might appear. This essay describes how the level of consensus on a topic can be assessed.

Why is the concept of levels of consensus important?
Many dispute resolution (and enforcement / disciplinary) mechanisms will consider whether someone is acting against a consensus held among other editors. With the exception of a few legally derived or WMF-mandated standards, the way an editor acts or the content they publish in whichever space they are working in, is judged against "this what everyone else has agreed we should do in these situations".

Due to the fact that the community has consisted of millions of usually anonymous people over many years, most of whom have never commented anywhere outside of article space, it is impossible to ever accurately state what "the community" thinks about any given topic at any given time. There are some clues though: informal agreements on talkpages and elsewhere, Requests For Comment, topic-oriented noticeboards, and dedicated policy and related discussion pages.

It is considered disruptive to knowingly act against a previously held consensus, but how to manage situations where consensus may have changed, or different groups have different opinions on whether a wide consensus even exists? The views of one subset of the community's views are no more or less valid than another, and anyone can ignore the rules at any given point if required, but the one of the ways the community (or at least the part of the community that wishes to get involved) judges whether or not a new consensus is emerging is to contrast the level of new consensus against the old. This approach helps to prevent old and possibly contentious issues being constantly re-litigated, but is not without it's own challenges.

Defining some terms
CONLEVEL states "Consensus among a limited group of editors, at one place and time, cannot override community consensus on a wider scale."

Exactly what "limited group", "one place and time", "community consensus" and "wide scale" mean is open for interpretation. At least in the context of this essay, it's important to establish some terminology.

Topic The subject under discussion. This could be article content, or the application of policy.

Discussion The interactions between at least two editors concerning the subject.

Consensus The outcome or agreement arising from the discussion regarding the topic.

Local / Global Relative terms that indicate whether the outcome is more or less limited in application than a different discussion on the topic.

Widely / Narrowly Advertised Indicates the volume of editors who can reasonably be expected to be aware of a discussion and any subsequent consensus formed.

Correctly / Incorrectly Advertised The degree to which interested editors who might expect to be informed and who will be impacted by the outcome can reasonably be expected to be aware of the discussion.

Some common, but not universally accepted, assumptions and related caveats
Policy text represents the highest level of community consensus on a topic. This is often true. One of the reasons we have policies, and the expectation that they are followed in the absence of a compelling reason, is that they document established consensus so that it can be referred to. It is important to note that the policy text only represents the consensus that led to it being added. Policy pages are often watchlisted by editors, which means that unrepresentative or malicious changes tend to be reverted, and even minor alterations are often analysed in great detail on their talk pages. Not every change is subject to a formal or advertised discussion, and so in contrast to the number of editors in the community, only a relatively proportion of them may even be aware of the change. Also, by it’s nature, policy text reflects the thinking of editors at the time was written; it seldom leads the way, and it can become outdated as editing practices changed.

Guidelines represent the highest level of community consensus on a topic. This is often true. Guidelines usually give more detailed guidance on how policies should be applied, or as in the case of the MOS, provide community with best practices that should generally applied during editing. However, the caveats mentioned with respect to policies also apply to guidelines. The page watchers may be fewer, especially for the more specialized or older pages, so there is less community oversight of changes.

If a discussion is at the wrong venue, then any resulting consensus is more local than that from a discussion at the correct venue. This can be true. Certainly being at the wrong venue may negatively impact how widely (how many people have the page watchlisted?) and correctly (are the people who can meaningfully contribute to a discussion on the topic aware?) the discussion is advertised. But these issues can be mitigated to an extent with sensible notifications, and discussions at talk pages have attracted hundreds of contributors on occasion.

If a discussion is not a properly formatted RfC, then any resulting consensus is more local than from a discussion that is. This can be true. An RfC generates automatic notifications (for those who have signed up for them) and is theoretically visible to the whole project via central listings. Also, the discussion format helps to structure and regulate the discussion in way that leads to clearer outcomes and (sometimes) a formal closure by someone uninvolved. There are no guarantees though that the “right” people see the notifications at the right time, or indeed attend.

A discussion with a high number of attendees means the consensus formed is global. This is often true. In the situation where 100% of active editors discussed something, the outcome would be a truly global consensus. However, even the most well-attended discussions generally only attract a fraction of the community, and below a certain level this ceases to be meaningful. Does a discussion that involves 4 editors have a higher consensus level than one involving 3? Exactly where the line lies is an open question.

The outcome from an discussion should be applied in other similar situations. This can be true. If there is consensus to include something in policy, then it should generally be applied during editing. If there was another form of discussion on a topic, it should also be applied for the topic in question. However, reasonable minds may differ on exactly whether the level consensus reached in a particular discussion means that this should ever apply to anything else in the future. A discussion outcome is not always a mandate to treat everything that is similar but not identical in the same way, although it can be a useful guide.

A consensus can exist without there having been an explicit discussion. This can be true. A silent consensus arises when edits are made and continues until the edit is modified, challenged, or reverted. This can also be said to apply to content which has gone through a peer review process, e.g. FAC. An unchallenged edit on much-watched page may be presumed to have a greater consensus than an edit made relatively recently on a little-watched page. In either case, the fact that the original edit was not immediately challenged does not provide a substantive basis to support keeing it.

So what does this mean?
It means that there is no simple formula for determining the level of consensus within the Wikipedia community, and this should always be borne in mind when discussing with others. There are some commonly accepted elements that can be considered as increasing the level of consensus associated with a discussion outcome, but there no consensus on how to weigh them in proportion each other.
 * Already documented in policy or guideline
 * High number of editors aware of discussion
 * Appropriate central noticeboard
 * Formal RfC
 * Correct venues / participants notified of discussion
 * High number of participants involved
 * Explicitly wide scope identified

All of these elements should be considered together when trying to determine the level of community consensus there is on something. So what recommendations are there in terms of how to frame or assess arguments in relation to consensus level?


 * Do not dismiss something as being "only a local consensus" or use disparaging terms without explanation.
 * Do not baldly state that there is already a “global consensus” on something without referring to both the discussion and why you believe it carries more weight than something else.
 * Do carefully check the scope of what was decided in a discussion. Just because there was a recent discussion that concluded that we should always handle Foo a certain way, doesn't mean the same should apply to Bar. Even if it appears clear to you that it should be applicable, be open to discussing further.
 * Do remember that even though you participated in a discussion that occurred on a page you consider to be important to the project, that anyone who challenges the applicability of that outcome elsewhere might have legitimate reasons to do so. Other editors may place more or less weight than you on the number of actual participants, or the venue in which it took place.
 * Do follow policies and guidelines wherever possible, but when someone wants to do things differently, remember that Wikipedia is always open to change, and that seeking change can be a positive thing for the encyclopedia and readers.
 * Do not endlessly challenge or refuse to accept any form of consensus formed elsewhere. Although consensus levels are subjective, it's disruptive and contrary to the philosophy of the project to reject all forms of consensus for not being "global" enough.