User:Tgr (WMF)/blame maps

From mediawiki.org

Wikipedia prides itself on its transparency, yet for the average reader it is completely unfeasible to trace the provenance of an article, and even for one familiar with Wikipedia/MediaWiki it can be a challange. Wikipedians rely on treating articles a change feeds, because single edits are relatively easy to grasp; once something gets through the filters of recent changes / watchlist, it is lost in the sea of past history. This makes the review workflow somewhat ineffective as the sum of the sizes of edits can be much larger than the size of the article. So both reader trust and editor effectiveness suffers.

Article with more recent additions highlighted in darker color, as rendered by the (now defunct) experimental tool WikiTrust

The likely solution has long been identified (and had some experimental implementations): maintain a map of authorship data, something like git blame (or rather, git blame --word-diff), and than use that to annotate pieces of the article text with authorship, or age, or some trust metric calculated on top of the authorship map.

Besides establishing reader trust (or un-establishing where appropriate) and providing a tool for journalists and researchers that's much more effective than page histories, this could also support less onerous review workflows for experts with limited time (visit the article I have written every three months and see the changes highlighted, with authorship data, maybe the ability to set up a trusted / suspicious user list) and more granular notifications (tell me when text I have written has changed / open a talkpage section about the hightlighted sentence and ping the editors who have contributed to it).

See also: