Andrew Gray wrote:
First off, you need to quantify new content added rather than, say, edit wars; I believe there are existing algorithms to manage this. Filter out only those content-adding posts - maybe 50%, maybe 5% - and then check to see if that content still exists in the final version.
How can you efficiently do that? -We have article at revision A -Anon vandalises article -Anon adds content needed to be wikified. -User reverts to version A -User wikifies and rewrites anon's text.
A machine check would attribute the last addition solely to the user, while it's a derivative of the anonymous one, even if it's clearly stated on the summary.