Community Insights/2018 Report/Anti-Harassment Tools

Translate this page

The mission of the Anti-Harassment Tools team is to help Wikimedia volunteer communities reduce the level of harassment and disruptive behavior on our projects using research, product development, and policy growth. The Anti-Harassment Tools team is part of the cross-departmental program called Community Health Initiative, which involves people from both the Contributors team and Community Engagement department.

For this survey, the Anti-Harassment tools team would like to answer the following questions:

What are the characteristics of contributors on Wikipedia who try to resolve conflicts?
Why do contributors help to resolve conflict?
What is the climate for free speech on the Wikimedia projects?

Results

1. What are the characteristics of contributors on Wikipedia who try to resolve conflicts?

Among contributors on the projects, 43% reported having tried to resolve an argument or conflict between other users on any Wikimedia project (AH01). Upon examining how contributors at different activity levels contributed to this question, we find some differences between groups. In general, contributors with higher activity levels were more likely to report having tried to resolve a conflict.

**Table 1:** Percent of contributors who reported intervening in conflicts, by Wikimedia project
Project	Yes	Total n
Wikidata	0.00%	6
Spanish Wikipedia	14.30%	7
Arabic Wikipedia	26.70%	15
French Wikipedia	30.90%	55
Italian Wikipedia	31.30%	16
Russian Wikipedia	35.30%	17
MENA Wikipedias	36.40%	11
Portuguese Wikipedias	36.40%	11
CEE Wikipedias	38.50%	13
Other Wikimedia projects	38.90%	36
English Wikipedia	42.90%	14
Dutch Wikipedia	44.70%	38
Western Europe Wikipedias	46.00%	63
Chinese Wikipedia	49.20%	59
Commons	52.60%	19
German Wikipedia	53.80%	65
Asian Wikipedias	58.30%	12
Japanese Wikipedia	68.40%	19

When exploring this question by Wikimedia project, respondents from Asian language Wikipedias seemed more likely to report intervening in conflicts. Response rates by project vary widely for this question and many had low response rates. Taking this into consideration, we see some possible trends. Upon categorizing the data by interquartile range in Table 1,^[1] the data suggest that Wikipedians from Asian languages (Chinese*, Japanese, and Other Asian languages) were more likely to report trying to resolve conflict, while respondents from French*, Italian, and Arabic Wikipedia seem less likely. (* = n above 30 respondents).

It's important to note that these are associations only, and we are unable to identify confounding factors due to the low numbers. For example, we are unable to calculate whether certain projects have a higher proportion of very active editors, which might influence the results we see in the table. We also do not have data on how much conflict exists in a particular project, which might also increase the opportunity to intervene in a conflict.

"If I think I can help people get along, that is contributing to Wikipedia just like any other contribution."

—Wikipedia editor

2. Why do contributors help to resolve conflict?

The majority of sampled respondents reported helping to resolve conflicts for the sake of Wikipedia (AH02). Within the responses, there were two major themes.

Helping Wikipedia as an encyclopedia (38% of 120)
Helping Wikipedia as a community (39% of 120)

Within the encyclopedia theme some examples include the following. Adhering Wikipedia policy, keeping the focus on writing content, and conflict being harmful for the project. Within the community theme, statements revolved around helping individuals or for the greater purpose of keeping the community moving forward. Less prominent themes included; the perceptions that someone needed to help (12%), conflict is a waste of time for everyone (11%), and empathy for the victim (10%).

The smallest relevant themes were because the conflict touched on personal values around justice, for personal reasons, and because they had information that would help. A theme surfaced that participants intervened out of duty as administrators. This theme was found to be part of the unsampled participants. It is important to note that the phrasing of this question may have introduced bias in responses. The question included the word "help", which may have elicited sentiments of altruism from respondents.

3. What is the climate of free speech on WIkipedia?

Results from CE Insights 2018 report

Contributors were asked a few questions about their attitudes toward free speech on Wikipedia, and whether they felt they were able to freely express their thoughts (AH03). These questions were asked on a scale from 1 to 5, where 1 was "Strongly Disagree" and 5 was "Strongly Agree".

65% (n = 453) agreed or strongly agreed that their freedom to speak up is protected on Wikipedia. On average, women reported statistically significant lower scores than men. The mean response for females was 3.2 and for males was 3.7 (both between "(3) Neither Agree nor Disagree" and "(4) Agree").^[2] We found no differences when comparing different regions and different activity levels.

54% (n = 447) agreed or strongly agreed that they are freely able to express their thoughts without being attacked on Wikipedia. On average, women reported statistically significant lower scores than men. The average response from female participants was 2.66 (between "Disagree" and "Neither Agree nor Disagree") and was significantly lower than males, who reported an average of 3.43 (between "(3) Neither Agree nor Disagree" and "(4) Agree").^[3] We found no differences when comparing different regions and different activity levels.

45% (n = 428) agreed or strongly agreed that they know where to turn to for help when someone was verbally attacking them on Wikipedia. On average, respondents from emerging countries and editors with high activity reported statistically significant higher scores. Respondents from emerging countries reported an average of 3.40. This was higher than respondents from globally Northern countries. Respondents from the Global North reported an average of 3.06 (both between "(3) Neither Agree nor Disagree" and "(4) Agree").^[4] When looking at global regions, we detected no differences. The average response for low-activity editors was 2.89 (between "(2) Disagree" and "(3) Neither Agree nor Disagree"). This was lower than high-activity editors. High-activity editors on average reported 3.28 (between "(3) neither agree nor disagree" and "(4) agree").^[5] No differences were found in how people from different genders responded to the question.

Most useful findings

It seems that women editors felt less able to express their opinions on wiki without being attacked.
Some languages show different levels of willingness/attraction to mitigate conflict than others.
From the users in the survey, less that half agreed or strongly agreed that they know where to turn to for help when someone was attacking them with words on Wikipedia.
There appears to be increasing levels of conflict mitigation as accounts are more active. This suggests that heavier engagement correlates to increasing desire to mitigate conflicts and/or higher exposure to conflicts.
There seems to be an even split between motivations for mitigating conflicts. The health/quality of the encyclopedia, and the health of the community. This illustrates a possible divide in interests among contributors between the content and the people.

Next steps

In the coming months, there will be conversations about Community Health Metrics. We would like to ensure that we add community health metrics to the CE Insights survey in the future.
This information will inform discussions we are having about a better reporting system for conflicts.
This data can inform where we test new approaches to supporting users who are going through conflict.
It is good to know that women responded differently to these questions about not being able to freely speak up. This can help inform conversations and discussions.
There is a research project around monitoring cases of incivility. This data will inform those discussions.

Notes

↑ https://en.wikipedia.org/wiki/Interquartile_range
↑ Means were reported, Mann-Whitney test, p-value = 0.002; effect size was small: 0.15
↑ Means were reported, Mann-Whitney test, p-value < 0.001; effect size was small: 0.18
↑ Means were reported, Mann-Whitney test, p-value = 0.035; effect size was small: 0.10
↑ Means were reported, Mann-Whitney test, p-value < 0.001; effect size was small: 0.17

[1] ttps://en.wikipedia.org/wiki/Interquartile_range

[2] Means were reported, Mann-Whitney test, p-value = 0.002; effect size was small: 0.15

[3] Means were reported, Mann-Whitney test, p-value < 0.001; effect size was small: 0.18

[4] Means were reported, Mann-Whitney test, p-value = 0.035; effect size was small: 0.10

[5] Means were reported, Mann-Whitney test, p-value < 0.001; effect size was small: 0.17

[1]

[2]

[3]

[4]

[5]