Some good questions here by Steve Bennett, which deserve good answers. I'll try:
a) Most people will abandon "nn, delete" reasoning and seriously consider
good
arguments for inclusion if someone throws a strong argument into the debate.
What if the keep argument is made at the end of the voting period? It's obviously the great weakness of AfD that discussion and voting happens simultaneously - votes can take place in ignorance of intelligent arguments subsequently made.
20 straight "nn delete" votes followed by 3 straight well argued "keep" votes is something that I for one would take into consideration. If the admin chooses to close as a "keep", I don't know if DRV would be able to muster the required 75% majority to "overturn and delete", if it does the article *was* maybe something which ought to be deleted after all, the experts' opinions notwithstanding. A 50% majority to overturn will only produce a fresh relisting, with the keep argument getting the attention it deserves, some more discussion, and a more thought out result.
Alright, let's say the admin chooses to delete. Sometimes those who vote on DRV will be sympathetic to the argument that the last three contributions to the discussion were not adequately addressed because they came so late. Perhaps especially if the last "voters" made some considerable improvements to the article shortly before it was deleted.
Well, sometimes not, I do agree that "Keep deleted. Valid AFD debate. ~~~~" can be a real pain to those who make a good effort to save an article which arguably should not have been deleted. It is a problem, not one that pops up very very often, but still annoying at times.
b) We have a number of inclusionists on AFD who more or less reject the
notion
of notability anyway and will vote to keep articles on all roads, streets, schools and churches.
I think it would be in the project's interest if we could define a set of exceptions to "notability" on the basis that comprehensiveness in certain areas is more valuable. Most people would probably agree that every university in the world should have an article. However by definition, once you include "every" anything, you include "non notable" examples.
In other words, I don't think every subject should have to be notable, if it has another reason for being included.
In some cases we do have a consensus. We automatically include all articles on any thorp, hamlet or village, even if there only live 5 people there. A few AFD debates have pretty much determined that all railway stations and subway stations are not subject to notability requirements. When the Hippopotamus Defence article was kept, it pretty much determined that all chess openings are worthy of articles, no matter how obscure they are (although I and some others have boldly gone ahead with some merging the subvariations of various openings).
On many other things, we will probably never form a consensus but all is not lost since we can learn by trial and experience. We don't have a consensus to keep all verifiable schools, but most have acknowledged the futility of trying to delete them.
c) Most people are loathe to delete well-written articles, even if the notability is dubious.
That's a major problem. Perhaps if we change the emphasis from "deleting it", with its connotations of purging it from the surface of the earth, to "moving it out of Wikipedia"?
Not really a major problem. Well written articles on borderline notable subjects don't really do much harm to the project. At best such articles illustrate the depth and vastness of information which lies in Wikipedia, and as long as they are verifiable, neutral and tertiary, they don't really do much harm.
"Moving out of Wikipedia" is a term I would reserve for transwikis.
Sigvat