Wiktionary talk:Votes/pl-2015-07/Disallowing extending of votes
Rationale[edit]
Maybe this (not my rationale): extensions of votes allow the extender unfair influence on the outcome of the vote. See more at Wiktionary:Beer_parlour/2015/July#Persistent_extensions_of_votes. Supporters, please provide your actual rationale. --Dan Polansky (talk) 09:01, 19 July 2015 (UTC)
- My objection: we need to allow extensions of votes since people are slow to come to votes, and the 2/3 threshold for pass is really challenging to meet. Admittedly, there is a very slight influence that the extender has on the outcome of the vote by choosing when to no longer extend the vote. However, another extender can extend the vote, then. More importantly, the longer the vote stays, the more representative it is of the consensus of the whole editor community as opposed to the fraction of the community that initially comes to the vote, which plentifully offsets this disadvantage. --Dan Polansky (talk) 09:08, 19 July 2015 (UTC)
- Responses to the objection: Re "since people are slow to come to votes": then have a longer time from the start. Re "another extender can extend the vote, then": if he happens to catch it before someone closes it. Re "the longer the vote stays, the more representative it is of the consensus": except that repeated extensions followed by closure as soon as a certain percentage is reached means it's representative of that percentage; note that [[w:selection bias]] gives "[e]arly termination of a trial at a time when its results support a desired conclusion" as an example of selection bias (and deciding when to stop repeated extensions amounts to the same thing).—msh210℠ (talk) 04:09, 20 July 2015 (UTC)
- Re: "then have a longer time from the start": That would force even easily passing votes to linger more than the minimum of one month. That, IMHO, is undesirable. It is, nonetheless, desirable to give votes that may be a pass a chance by extending them, since for them, the extension does not slow down their execution (they would be "no consensus" otherwise), and makes them more representative.
- Re: "repeated extensions followed by closure as soon as a certain percentage is reached": yes, I objected to that closure, and was prevented from posting to the vote page by the closer who locked the vote page. That is, I oppose the "as soon as" closure when it is a day after the threshold-surpassing vote was cast, but I think it is okay when the closure takes place one month after the threshold-surpassing vote was cast.
- Re: Yes, there is some selection bias, but very minimal one. It can be minimized by preventing the type of closure that you have criticized. The risk of that minimal selection bias is worth it, IMHO. --Dan Polansky (talk) 18:33, 20 July 2015 (UTC)
- Examples of easily passing votes are Wiktionary:Votes/pl-2013-10/Removing SAMPA and X-SAMPA, Wiktionary:Votes/pl-2010-05/Names of specific entities or Wiktionary:Votes/pl-2012-02/Brand names and physical product 2. --Dan Polansky (talk) 18:39, 20 July 2015 (UTC)
- Responses to the objection: Re "since people are slow to come to votes": then have a longer time from the start. Re "another extender can extend the vote, then": if he happens to catch it before someone closes it. Re "the longer the vote stays, the more representative it is of the consensus": except that repeated extensions followed by closure as soon as a certain percentage is reached means it's representative of that percentage; note that [[w:selection bias]] gives "[e]arly termination of a trial at a time when its results support a desired conclusion" as an example of selection bias (and deciding when to stop repeated extensions amounts to the same thing).—msh210℠ (talk) 04:09, 20 July 2015 (UTC)
Relatedly, I think the following is a good principle:
- If a vote is getting the result of no consensus, and if at least one vote was cast to it during the last month, the vote can be extended in good conscience. A result is "no consensus" if it is 50% in support or higher but lower than a pass.
This principle does show a slight selection bias, but not one that we should be concerned with, IHMO. It helps more votes reach a definite result rather than ending in the indefinite "no consensus". It does not make votes showing a clear result linger for months, only those that do not show a clear result. --Dan Polansky (talk) 18:25, 21 July 2015 (UTC)
- @Dan Polansky: Hmm. That's a rather good idea. — I.S.M.E.T.A. 18:38, 21 July 2015 (UTC)
To reduce selection bias of the above principle of mine, I came up with the following extension mechanism:
- If a vote gets a clear result in the original unextended period, with enough participants, it gets closed with that result. Clear result is either a clear fail at less than 1/2 support or a clear pass with, usually, 2/3 support.
- Otherwise, the vote can be extended by a month.
- For a vote extended at least once: If new votes were added to the vote in the last extension period, the vote is automatically extended by one more month, regardless of the result at that point. A vote switched is not a vote added.
The above is rather stringent, maybe excessively so to my taste. And it still has a trace of selection bias since it distinguishes an early pass from a late pass. The only way to remove that remnant of bias is to never extend a vote, which I think is inferior. I surmise that, given the relatively small number of editors we have, extended votes show a strong tendency of being more representative, which offsets the bias. It appears to me to be very much unlike some of the reported malpractices of big pharmaceutical companies during drug trials. --Dan Polansky (talk) 07:14, 8 August 2015 (UTC)