Most try to keep private various categories of personal information (p8)
40% of respondents contribute more than once a day ; 8% no longer contribute to Wikimedia (p9)
80-85% of respondents identify as male (p11)
40% have graduate education ; average age about 35 (p12)
45% edit principally in English or that is their native language (p13)
Sizable numbers have experienced pretty extreme harassment, e.g. stalking, revenge porn, hacking, impersonation, threats of violence -- number of cases was not clear or was ambiguous, but 1215 respondents answered the question and number of implied cases might have been either on the order of 2400 (2 cases per respondent) or 24 (2% of respondents). (p19)
In a fuller written report, explicating that would be useful and should be made clear.
Respondent counts (the denominator) and counts of cases unclear, pp 22-24 and p25 ; it looks like there's a huge amount of harassment from the numbers presented. If I were a discussant of these findings, I'd emphasize that this needs to be clearer.'
Some illustrative nasty quotes and descriptions, p27-28. 'It could be useful to gather the worst cases into a list so we have for example a list of written death threats on Wikimedia -- It is thinkable that these per se would be grounds for some kind of expulsion or criminal complaint.
Response to experienced harassment is not usually effective but in a substantial minority it has some effect (p32)
Female users apparently much more likely to report incidents to the Foundation (p30-31)
Effectiveness of response to harassment, differentiated by type of response, is interesting but hard to summarize or draw many lessons from. Note however that asking for lawyers or law enforcement to act appears to have a really low rate of success. (p33)
Half or more of respondents had seen some extreme form of harassment on-wiki.' Number of cases not clear, here, but probably number of respondents to these questions was over a thousand. Many will have observed the same key csaes. It is implied that there are thousands of observations of extreme harassment, though number of cases might be 100x smaller. Hmm. Actionable implication: quick effective response to even a few extreme cases could cause a large change in the perception of the platforms, since so many people would either observe less harassment or observe effective response to it. (p35)
People experiencing harassment reduced their involvement on Wikimedia, and those observing harassment tended to participate somewhat less afterward. (p43-44)
Useful suggestions from respondents, p47. Too little is shown here. Useful definitions and classifications, pp 48-50.
Emphasis: 36% of respondents to this question said they'd obsessed threats of violence on-Wikmediai. Respondents to this question probably numbered 1215 or more. (Denominator is not clear) (p35) My instinctive response is that this single data point guides us to action, and a civility task force would help and a pretty specific effort by the WMF to contact the harasser and law enforcement would be a good complement.
Resolution paths are discussed a bit on pp 40-41. A clear description of these paths and a to-do list for people who experience harassment or their aides could be helpful; presumably like me many would feel afraid and not know what to do.
Suggestions for improvement included idea of karma points get accumulated, like edit count. Trolls and other bad apples could be identified by scores deriving from multiple people. Let users protect their own talk pages from various categories of other editors. Admins, ANI, ArbCom should act more quickly. WMF should offer a clear path to reporting harassment on email or IRC. Train admins more. Put finite terms on admins -- I guess this addresses problematic admins, although that problem was not diagnosed in the report. Tell new users in Welcome message how to report problems. Focused comments on blocking users more effectively by IP address.
Findings about on-wiki threats of physical violence (above) seem very actionable -- at least there could be a central list of these. Adding an item to that list could be very decentralized so gnomes could do it, perhaps anonymously, with hyperlinks to the threat. The list would never be complete, and there are always ambiguities. This seems really doable, even by me alone.