Last modified: 2014-08-15 12:01:56 UTC
This proposed new feature would automatically filter posts from users who consistently add comments that are inappropriate. The goal of this feature is to reduce the workload for editors who monitor the feedback page, by filtering out new posts from users whose previous posts were found unhelpful, flagged for abuse, hidden and/or oversighted. One way to implement this feature would be to average the relevance score of each user's previous posts, and storing that value in a data table associated with that user's ID or IP address. If a user's average relevance score is negative, their posts would be removed from the 'Most Relevant' filtered list (alternatively, they could be auto-flagged, but without subtracting '-5' points to their already low relevance score). If a user's relevance score is positive, no filtering action would be taken, and their posts would appear as they do now. This proposed feature is likely to have a significant impact on database performance, so we will investigate it some more before implementing it. But if we can make it work, this could substantially reduce the editor's workload, which is a key objective for us at this time. To learn more about relevance filters, check our feature requirements page: http://www.mediawiki.org/wiki/Article_feedback/Version_5/Feature_Requirements#Relevance_filter
The first action item for this feature is to determine its feasibility from a technical standpoint: it is likely to require another database field, which may impact our database sharding plans. A variation on the proposal above would be to track users based on their unique personal token, rather than their IP address. The reasoning is that this token (now stored by AFT in a cookie) may be a closer match for that user than the IP address, which could be shared by many different users, or change dynamically. The downside is that one user may have several different tokens, one for each browser and device they use. Let's investigate the feasibility a bit more, then discuss our next steps for this feature based on our findings. Thanks.
[Lowering priority to reflect reality, as AFTv5 is not very actively being worked on anymore.]
