Topic on Talk:Edit Review Improvements/New filters for edit review

Most of these were not good, such as:

https://en.wikipedia.org/w/index.php?title=List_of_countries_by_refugee_population&curid=24201228&diff=810622370&oldid=808827193

https://en.wikipedia.org/w/index.php?title=2010_Chile_earthquake&curid=26357457&diff=812661570&oldid=812660978

The filters are predictions, so you may have false positive (check the diagram on that page).

How many changes have you reviewed to get those two edits?

Hi, unsigned user. Thanks for asking about these filters. If you want to exclude most or all of the bad results from your "Very likely good" searches, you can achieve that I think by also selecting the "Very likely good faith" filter. This way, you'll get only those results that are both good and good faith. Like so:

https://en.wikipedia.org/wiki/Special:RecentChanges?damaging=likelygood&goodfaith=likelygood&hidebots=1&hidecategorization=1&hideWikibase=1&hidelog=1&limit=250&days=7&urlversion=2

If you want to refine your prediction even more, set some highlights (without checking the boxes to activate the filters) for "May have problems" and "May be bad faith." That will let you visually pick out the edits that are more doubtful. Like so:

https://en.wikipedia.org/wiki/Special:RecentChanges?damaging=likelygood&goodfaith=likelygood&hidebots=1&hidecategorization=1&hideWikibase=1&hidelog=1&limit=250&days=7&damaging__maybebad_color=c3&goodfaith__maybebad_color=c4&urlversion=2

As Trizek says, the "User Intent" and "Edit Quality" filters produce probabilistic predictions. Meaning they're not always right, but the likelihood is good. For both Very Likely good" and "Very likely good faith," the false positives should be pretty infrequent—one or two % according to tests. But we'll be very interested in hearing more about your experiences—let us know if these filters are working for you, and whether those tips help.