Wikipedia:How to hold a consensus vote

I'm withdrawing this proposed guideline. I created it at a time when a whole bunch of surveys in a row were being botched. But I've decided this isn't really necessary, especially the part with the overly-detailed discussion of various voting methods; Straw polls should be enough.  r  speer 17:46, 4 November 2005 (UTC)

There are times in a Wikipedia discussion when you want to hold a vote, to encourage people to give their opinion so that you can find a consensus.

It is useful to look at the Wikipedia articles on Voting systems before organizing the vote; however, you should keep in mind that you're looking for a consensus, not a decision, so your needs will be very different from the assumptions made by those articles. The consensus should always be more important than the vote, and you have to take this into account. You can't use the same system used to elect, say, the President of Ireland, to reach a Wikipedia-style consensus.

So here are some suggestions on how to organize a vote so that you achieve a consensus and avoid a fiasco.

First choice: Don't vote, just discuss
Is it at all possible to reach a consensus without voting? This is a preferable option for everyone. Remember that Wikipedia is not an experiment in democracy, and throwing in a democratic procedure for the hell of it won't accomplish anything useful.

You may have recently read about some voting method that you think is really cool, and you want to try it out. Don't use Wikipedia to try it out. There are other places where you could do that more successfully.

In most situations, the most effective way to reach a consensus is through discussion. This discussion could involve a straw poll, with section headings under which people can comment and give their support for each option. This may even look like a vote, but you don't determine the consensus by counting people's comments, but by reading them.

This kind of structured but informal discussion fits well with the Wiki editing style. People can create new options to support, and go back and change their comments on other options in light of the new options. Often a consensus can be found when someone suggests a new option that is acceptable to everyone.

If you go on past this section and decide to hold a vote, it most likely should be because a discussion was already tried and was inconclusive.

Second choice: Find two clear options to choose between
As soon as you have more than two options, you get into dangerous territory. You'll have to choose between plurality, approval, and preferential voting, and no matter what, someone will be upset by the choice.

On the other hand, everyone knows what to do when there are two options. They'll focus on supporting one option or the other, or (hopefully not, but possibly) telling you that the options suck and you should hold a different vote, but the attention won't be on how you're such a fascist for imposing your voting method on the decision.

Often, a proposal will be a short description, and the two options are to accept it or oppose it. In this case, it should be obvious that a majority or consensus to oppose the proposal does not imply majority of consensus to accept its opposite. As an example, if the proposal is for a policy that "All Wikipedians must wear a hat" and it gets overwhelming opposition, that does not imply that a policy of "Wikipedians may not wear a hat" is now accepted.

Third choice: Approval voting
You can't avoid it – you've got more than two options, and most of them aren't going away.

At this point, you should keep these things in mind:
 * If any person misunderstands any part of the method, and convinces other people that their misinterpretation is right, you're doomed. You will have a fiasco, and not a consensus. Your first priority, then, should be to choose a method that is so simply expressed that it can not be misunderstood.
 * As a corollary, if someone is malicious and supports the status quo, they may deliberately spread a misconception about the vote, and therefore cause consensus to fail. The method has to be so simple that such a person will be obviously wrong the moment they speak up.
 * However, if you use plain plurality voting (vote for one) on three or more options, you probably won't get a consensus. You probably won't even get a majority. So you can't get this simple.

Approval voting is simple, easy to implement with Wikipedia's formatting, and will work in many situations. You should explain it something like this:
 * Here are the options. Vote for as many as you want. The option with the most votes is the consensus.

Optional addition, if the issue is contentious:
 * If no option is supported by over x% of the voters, there is no consensus.

Do not set x too high, or you'll end up needing another vote. 50% is reasonable. 75% very rarely occurs in approval voting, due to people only voting for one option even when others are acceptable to them. 60% is a pretty strong consensus. However, multiple options may achieve strong or relatively equal approval, in which case consensus is likely to be challenged.

And don't say "over x% of the votes" - people will interpret this to mean that voting for three options is three votes, and then it won't even be possible to get a consensus.

One major downside to Approval voting is that it encourages voting blocs to express approval for one and only one option. Those who express approval for more than one option are liable to be discounted, if the one-and-only-one blocs outnumber the moderates. In contentious polls, this swings the result to the extremes rather than the most generally acceptable alternative.

Now, Approval looks too simple, and the temptation will be to add on other stipulations, like:
 * Let's hold a runoff vote between all the options that get over 50% / the top two options / etc.
 * Let's make it possible to cast a disapproval vote, too.

Do not do these, or anything else of the sort. Not only are you making up a new method that isn't Approval and won't work as well, but you are adding new rules. These rules may backfire, causing a result that is clearly not the consensus, and they are likely to be misinterpreted.

Case study
The Template standardisation vote was held using Approval, but with the stipulation that if no option got over 75% support, there would be a runoff between the top two options.

The clearly favored winner got between 74% and 76% support, depending on which dubious votes you counted. The second place choice got around 30%.

The person whose template got second place calculated the support as not being over 75%, and started setting up the runoff vote. Other users saw this as him trying to put his option back in the running after he already lost, and made comments in the discussion that some users saw as personal attacks on the person who set up the runoff vote.

The issue was concluded with a consensus that the runoff was not necessary, along with some hard feelings.

If the template had been chosen with a straight Approval vote, with no extra conditions, the winner would have been clear and the voting system would have stayed comfortably in the background. However, had the second placed option received nearly as many votes as the winner, thus having two clear favorites, a runoff may have been appropriate.

Fifth Choice: Borda count
Borda count is more efficient after an elimination pre-round is done using Approval voting, taking all those options with at least an average score. Those higher approval options are then ranked according to the Borda count. However, Borda has been successfully used for creating a consensus without a pre-elimination. The Borda winner will have the highest overall ranked preference. This is usually the Condorcet winner, but not always.

Procedures for the vote
When you organize the vote, you should follow the procedures described at Survey guidelines, including reaching an agreement on the procedure beforehand.

If, during the period where the voting procedure is being discussed, lots of people tell you to simplify the procedure, they probably have a point. It is also likely that you will have various people telling you to complicate the procedure to deal with some special case. Try pointing them at this page.

Sanity-check your consensus
If you decided to hold a vote, it's because you decided that comments weren't enough for the situation, and you needed numerical data in the form of votes to find a consensus.

Even so, you should look at people's comments and take them into consideration. Is a "winner" emerging that is clearly not the consensus? Are people casting votes that seem irrational based on what they're saying? Are people complaining about not having the right options available?

In such cases, you may unfortunately need to draw some attention to the voting method, and look at where it's going wrong. You may not be able to consider the outcome of the vote to be a true consensus. Again, the consensus is more important than the vote.

Of course, your sanity check should not take the form of "I don't like this consensus, so it's clearly wrong." If you hold a vote and it comes to a clear outcome that you don't like, it goes without saying that you should stand by the consensus anyway.

Quorum
If your consensus decision is so momentous that you need a large number of people to participate in it, you'll want to set a quorum.

First of all, you are going to overestimate how many people are interested in this issue. In the Arbitration Committee proposed amendment ratification vote, they believed that there would be over 100 votes cast by people with over 500 edits. This was shown to be completely unrealistic. Keep the quorum reasonable.

The intuitive way to set a quorum is to say that the status quo remains unless at least n different people cast votes. See the Quorum article for why this is a bad idea: it induces a strategic advantage to not voting.

Here's how to set a sensible quorum:
 * If you're using plurality voting (such as because there are only two choices), say that there must be at least n votes for the option that is not the status quo, or else the status quo remains.
 * If you're using approval voting, say that the status quo will remain if no option gets more than n approval votes. (This n is a number of votes, not a percentage.)
 * If you're using Condorcet voting, say that the status quo will remain if the winning option is not preferred to the status quo by at least n voters.