On 8/23/06, Ruud Koot r.koot@students.uu.nl wrote:
If you can't distinguish between a small correction or vandalism, this probably means the fact isn't sourced. Either add a source, or if you can't at least a {{citeneeded}} or related tag.
Right. In my mind there are really two tightly related problems:
1) Catching that a simple piece of factual data has been changed. and 2) Actually verifying that the change is good.
(1) is more important, because if we don't know that a change is made we can't verify it... But I also think that (1) is the easier part to improve... We could have a bot detecting many such changes and place them on a list, for example. Not a complete solution, but with a little effort we could probably quadruple our effectiveness. Over time, such features could be enhanced in a decentralized way by things like invisible <fact></fact> tags that tell bots about immutable text. (bot would keep a list of fact tags it has seen in an article and warn if one is changed or removed)
Some cynical offlist mail I received suggested that I might really be advocating stable versions here.... and it's true that I think stable versions will be a vast improvement on this point. But even with stable versions I'm not confident that such changes would get the high level of scrutiny they deserve, and that any technical or social improvements we come up with would also be useful in a future with stable versions.
(2) is much harder... If the fact is cited to an online resource then it is just a question of motivation to get people to check. ... But what if it's cited to an offline resource or, more commonly, not cited at all?
I have mixed feelings about how to address (2)...
We could, for example, treat the case like a dispute over uncited material, where our default action is usually to remove the uncited claims and request that the disputing parties provide citations. But since there is so much totally uncited basic data, such a policy would have the effect of creating a new form of vandalism... and potentially result in an overall reduction of good information. We could also revert unless the later editor justifies their claim.... but I've seen a number of comment less date changes which did check out, so that too has problems.
I like the suggestion of changing the edit summary text to the most simplistic "Please explain your edit" or the like. I am not sure I see the benefits in mandatory edit summaries... at least today the lack of a summary is a reasonable indicator that the edit may need a deeper look.