Community Insights/2018 Report/Support & Safety

The Trust and Safety team identifies, builds and, when appropriate, staffs processes which keep our users safe. The team also designs, develops, and executes on a strategy that integrates legal, product, research, and learning & evaluation to proactively mitigate risk as well as to manage the overall safety of our online and offline communities when incidents happen. Trust and Safety comprises part of the Community Engagement department and works closely with the Legal department.

The Trust and Safety team would like to answer the following questions:

Are there changes to the level or types of reported harassment on the Wikimedia projects?
For user conduct policies on the Wikimedia projects, what is the level of awareness among users and perceptions toward their quality?
What does collaboration with the Wikimedia Foundation look like?
Are there changes to responding to support requests from Wikimedia communities?

Results

1. Are there changes to the level or types of reported harassment on the Wikimedia projects?

Across Wikimedia audiences, an average of 78% of 251 respondents reported never feeling unsafe or uncomfortable in any online or offline space in the last 12 months (CE16). In 2017 we asked the same question, but did not set a time limit of 12 months. In 2017, 68% of respondents reported having experienced harassment at some point in the past.

This possible decrease suggests there may have not been an increase in feelings of being unsafe or of harassment in the previous 12 months.

The most often selected choices for why people feel unsafe on the Wikimedia projects, were "Other" and "I do not feel unsafe" (CE01). The next most selected option was "Level of education" followed by race or ethnicity, and then gender. The least most selected options were disability or perceived disability and body type.

From those who felt unsafe, the areas where people selected "Sometimes", "Often" or "Almost always", were race or ethnicity (68%), gender expression (64%), body type (64%), and "Other" reasons (62%) (CE03). Least often reported in frequency were age and sexual orientation. For each of these "Sometimes", "Often", or "Almost always" were almost 50% of the respondents selections.

Some users of Wikimedia projects reported having avoided Wikimedia projects for one to three days because they felt uncomfortable or unsafe. Twice as many users avoided Wikimedia projects for one to three days, compared to other social media spaces like Twitter or Facebook (CE02). About 22% of Wikimedians reported avoiding Wikimedia projects for one to three days because they felt unsafe. About 10% reported avoiding Facebook or Twitter for one to three days.

Among the Wikimedia projects to which respondents contributed in the last 12 months, contributors reported a higher frequency of bullying or harassment on Wikipedia than on other projects. 29% of 280 reported having not been harassed on Wikipedia during the last year (CE05), while 79% to 96% of respondents said that they had not been harassed on Wikimedia Commons or other projects during the last year. The changes from 2017 were small.

When comparing to 2017 results, there were statistically significant changes for three projects; all between "Never" and "Rarely."

Wikimedia Commons (an increase from 1.14 to 1.25)
Wikiversity (a decrease from 1.05 to 1.0)
Meta (a decrease from 1.18 to 1.08)

When comparing across years by audiences, we detected no differences between high- and low-activity editors. The second project with the highest reported frequency of bullying or harassment during the last year is Wikimedia Commons. 21% reported a frequency higher than "Never". The project with the third highest frequency is Meta-Wiki, where 15% percent of contributors reported being bullied or harassed during the last year.

2. For user conduct policies and processes on the Wikimedia projects, what is the level of awareness among users and perceptions toward their quality?

Across various user conduct policies and processes, 40% reported that policies needed "Quite a bit" or "A lot" of improvement (SS08). For "Administrator selection and review processes", 38% reported they needed "Quite a bit" or "A lot" of improvement. 34% reported that "Blocking tools/mechanisms" and "Tools and processes for reporting users" needed improvement. Noticeboards had the lowest reported need for improvement at 26%. When comparing the data across audiences or across years, we detected no differences.

High-activity editors (VAE) know more than low-activity editors (AE) about user conduct policies on the Wikimedia project they spend the most time on (SS20). On average, high-activity contributors reported 2.9, or just about "Fairly well".

The average low-activity editor (AE) reported 2.44, closer to "Not very well". Based on a regression analysis, it seems editing activity is a better predictor of policy knowledge than the number of years of experience on the projects.

A majority of contributors agree that conduct policies are "sufficient and appropriate for their community" (SS21). About 67% of contributors agree the user conduct policies are "sufficient and appropriate for their community". 54% agree they are satisfied with the way user conduct policies are enforced.

Since contributors have different levels of knowledge of the conduct policies, their responses were checked for differences based on their level of knowledge. For the statement "I am satisfied with the way user conduct policies are enforced", no differences were found between levels of knowledge about policies. For the statement "The user conduct policies are sufficient and appropriate for the community," contributors that know the policies "fairly well" were more likely to report they agreed with the policies (Mean=3.6, n=220) than those who reported "very well" (Mean=4, n=71). This finding was statistically significant.

3. What does collaboration with the Wikimedia Foundation look like?

About 70% of 1,487 low-activity contributors and and 83% of high-activity contributors have heard of the Wikimedia Foundation. (SS14). About 23% of volunteer developers, 4% of program leaders, and 1% of Wikimedia affiliates reported not having heard of the Wikimedia Foundation.

In general, it appears that program organizers and developers have a stronger sense of the Foundation having an influence on their activities than contributors (SS23).

About 39% of all contributors said that the work of the Foundation influences their activities.

For developers and program organizers, about 65% found it true that the work of the Foundation influences their activities. Too few affiliate organizers responded to this question in order to report any results (n=11).

Developers reported knowing the most about how to contact the people they need at the Wikimedia Foundation (58% of 31). Low-activity editors (AE) knew the least (35%) (SS17). About 50% of program organizers reported that they knew how to contact people they need at the Wikimedia Foundation. 44% of high-activity contributors (VAE) reported having this knowledge.

No audience reported agreement that they are sufficiently consulted by the Wikimedia Foundation in decisions (SS17). Program organizers agreed the most (32% of 47) that they are consulted sufficiently by the Wikimedia Foundation. A greater proportion of program organizers (39% of 47) disagreed. The group with the highest disagreement response was developers. 47% "Disagreed" or "Strongly disagreed".

45% of both high and low-activity contributors disagreed that they are consulted sufficiently in their work by the Wikimedia Foundation.

No audience reported strong agreement that their voice was heard in Wikimedia Foundation decisions (SS17). About 41% of contributors, 39% of program organizers, and 32% of developers disagreed that their voice is heard in Wikimedia Foundation decisions.

4. Are there changes to responding to support requests from Wikimedia communities?

When asked where people would turn for help in an emergency, Wikimedians most often (48% of 100 sampled responses) reported they would turn to administrators or functionaries (SS05). About 24% reported they would go to law enforcement. 17% reported they would reach out to the Wikimedia Foundation. Three of 100 participants specifically mentioned the "emergency((@))wikimedia.org" email address. Other responses to this open-text question included communicating with the user or friends, reaching out to affiliates, doing nothing, reaching out to mental health professionals, and trying to do their best to help in the situation. This question went to all audiences.

In general, volunteer developers, program organizers and affiliate organizers reported having engaged with staff in the Wikimedia Foundation more than contributors (SS19). About 20% of high-activity editors reported having engaged with Foundation staff. 9% of low-activity editors reported having engaged with staff.

Contributors on the Wikimedia projects often reported contacting the Foundation for general help (SS03). About 52% of contributors contacted the Foundation for general help, while around 11% contacted the Foundation for emergencies, abuse investigation and dispute/abuse resolution.

Contributors most often reported that, when they have been harassed or attacked, they don't report the situation to anyone. The second most often place they report harassment or attacks to are to other volunteers (SS24). The third most often selected choice were funcationaries; such as bureaucrats. Contributors reported to the Code of Conduct committee, Wikimedia Foundation staff and the staff of chapters/affiliates the least.

Contributors found that getting support for a harassment or dispute from other volunteers or from Wikimedia Foundation staff was the most useful (CE13). About 29% found the Foundation "Mostly" or "Completely" useful. 28% found other volunteers useful. About 18% found Chapter/Affiliate staff and Functionaries useful. 13% found the Code of Conduct committee "Mostly" or "Completely" useful.

When reporting to other volunteers, the most contributors reported that problems were "Mostly" or "Completely" resolved (CE14). About 20% of those who reported to other volunteers about harassment or attacks responded that the problem was "Mostly" or "Completely" resolved. For the Wikimedia Foundation, the same measurement is 16%. For Functionaries and the Code of Conduct committee it was 15%. For staff of chapters or affiliates, 8% reported that harassment problems were "Mostly" or "Completely" resolved.

Most useful results

Harassment and bullying remains high on the largest Wikimedia project, Wikipedia.
The Wikimedia community does not feel sufficiently consulted by the Wikimedia Foundation.
We still do not understand the reasons why people feel unsafe.
Knowledge about what keeps people safe is lacking on many levels. Editors are unaware of policies and are unaware about the help offered by the Wikimedia Foundation. Meanwhile, the Foundation doesn't know about the problems users are having.
Volunteers (admins, functionaries) are important part of addressing harassment. They are in need of ongoing training and support.

Next steps

Use this report as an additional data point in driving more informed discussions around better reporting systems.
Improve communications channels to improve awareness of the emergency wikimedia.org email address.
Build a channel for non-emergencies through which we can be reached from our page on Meta-wiki.
Explore how to get better data from language-specific Wikipedia projects, through the Community Health Metrics kit and including measures into Community Engagement Insights survey.