Introduction on neutrality
The core Wikimedia policy is that the projects must maintain a neutral point of view. We are not allowed to take sides in any debate. This makes a good deal of sense, as firstly it means we don't risk offending people who have a different opinion from the one we present. Secondly it prevents us from simply making things up, as that would be presenting things from our point of view.
This goal is of course unachievable in practice, everyone sees the world from their own point of view. We have to hope that by meshing enough people's opinions we can achieve a balance that is neutral. As Wiktionarians make up a very small subset of the internet's population, it being a reasonably small subset of the total world's population, we cannot ever hope to have enough opinions to mesh into something neutral. When we couple the small sample size with the fact that we all share at least some interests (i.e. editing wikis) this gives us almost no hope for remaining neutral, there will always be some systematic bias.
This means that, instead of relying on what we think, we have to rely on what other people think. There are a good deal many more of them than there are of us, and they are more likely to be acting and thinking independently than we are. As we are a dictionary, the majority of the data that we provide is the meaning of words, either explained by definitions, or linked by translations or synonyms. Though we also peddle more specialist information, including Etymologies and the occasional usage note.
There are two places from which we could get the definition information: either we can copy it from other dictionary-like sources, or we can read written text and extract the meaning from the context. The problems with the first approach are many, not least it is likely to violate copyright law. It also suffers from the issue that we are preaching the point of view of that dictionary, though if we used enough dictionaries we would likely achieve a balance. Thirdly it means that Wiktionary can never exceed these other dictionaries, instead we would have to always remain subservient, snapping up new information as and when the other dictionaries add it.
This leads us to consider the other approach, reading text and extracting definitions therefrom. This has its own problems, not least the fact that there is a vast amount of text out there to read, and also that as we are trying to extract meanings from text, our ideas will spring from our personal experience and thus may not accurately reflect what was actually meant. As this method is still superior to copying other sources, we have developed the verification process which should help remove any mis-readings of the text, and the Citations: pages so that people can read the word in use and decide what was meant for themselves, completely independently of what we think.
The problems with descriptivism
Now we've decided that we are going to extract meaning from the text itself we are apparently faced with another problem. What if the people writing the text got it wrong? Maybe they misspelled a word, or used it in the wrong sense, they're only human after all. If we are not going to be relying on other dictionaries to get our information, what can we do?
If we stick fully to our idea of reading the meaning from the text, without looking anywhere else, then it seems that we should simply include these "mistakes" in an exactly equivalent fashion to the real entries that they are mistakes for, modulus a note that implies that this seems to be a less common way of writing the word, or a less common meaning of the word. This does indeed seem to be the logical way forward, however it makes sticklers and some philologists hold up their hands in horror. How on earth could you ever advocate that misspelling is acceptable?!
The only way to progress is to bring in some diplomacy and common sense. If we were to take the opinion that "misspellings are acceptable" this would definitely not be neutral, as it is not in accordance with what most of the world thinks. It makes sense if everyone spells things the same way, communication is less ambiguous. The problem for Wiktionary now becomes how do we decide who is right? We are not allowed to come up with our own opinions, so it seems that we must shelve our not derivative ideology and turn to other sources.
There are a few cases on which almost all sources will agree, misspellings are an example of this, and in cases like that it is acceptable for Wiktionary to adopt the same point of view as everyone else. Indeed, to choose a different position from the rest of the world would definitely be to state our opinions. In the large majority of cases however the other sources do not agree with each other. This is particularly the case with the usage notes that we provide, especially when you consider that most of our sources consist of written text that we are reading. It is not uncommon for writers to accidentally, or even deliberately, write in a way that grammarians consider unacceptable.
If the external sources do not agree with each other then Wiktionary is in a dilemma, it may not decide that one of the sources is more right that the other, that is not neutral even if one seems to have overwhelmingly more support, instead we must take a step back. By looking at the issue from our more distant perspective we can now report about the information given by the sources about the word. It is perfectly neutral to mention that "Some people think that.." even when it would not be neutral to say "It is true that...". In this way of avoiding being definitive, it might seem that Wiktionary is losing some of its usefulness, it is no longer proclaiming the rights and wrongs of usage. In fact the contrary is true, Wiktionary becomes more useful. By reporting that some people have one opinion, and that others may have another, Wiktionary allows the reader to decide for themselves which point of view is accurate.
As you will know if you've looked at our entry for descriptivism, the meaning of the word is very narrow and specific to linguistics. I would like to generalise this to imply that a descriptive source "describes" what it sees. By generalising this concept so that it covers describing not only the languages under discussion but also what other people say about those languages, we can include all the information that we would include if we were prescriptive and also maintain a relatively neutral position without being totally dependent on other dictionaries.