User:The Vintage Feminist/Antin et al graphics

"Gender Differences in Wikipedia Editing" by Antin et al, [1] describes the experiences of editors in the first three weeks of editing. This essay converts the results tables in their report into a set of at-a-glance graphics to provide a user-friendly way of considering their findings.

It strongly recommended, therefore, that this essay be read in conjuction with their report.

Background

edit

The report authors were interested in whether gender stereotyping rather than "active sexism" was the reason for the discrepency in the number of female editors compared with male editors on Wikipedia.[1]: 11–12 

Gendered Wiki-work

edit

The report considers the different types of "Wiki-work" that exists, from uploading photos to arbitration. It suggests that it is essential to discover which types of work draw which gender, what "are the real or imagined characteristics of tasks that encourage certain individuals to gravitate towards them", and to what extent stereotypes play a part. In order to grow Wikipedia's user base and encourage diversity, the report authors say that, researching wiki-work from a gender perspective "could allow recruitment and educational efforts to be more focused..." giving "potential contributors more specific ideas of the types of work they might like to do.[1]: 12 

Method

edit

Base population

edit

Figure 1: 256,190 new user accounts created between 9 September 2010 and 14 February 2011

  242,592 users who did not declare a gender in their Wikipedia profile (95%)
  13,598 users declared a gender in their Wikipedia profile (5%)

Between 9 September 2010 and 14 February 2011, 256,190 users created an account on the English-language Wikipedia. Of these 13,598 optionally declared a gender in their Wikipedia profile (blue segment, figure 1, right).[1]: 12 

The 13,598 gender-declaring Wikipedians were made up of, 11,194 (82%) men, and 2,402 (18%) women (figure 2, left).[1]: 12  A ratio that, the report authors' note, is higher than the one quoted in the 2010 UNU-MERIT study (13% women).[1]: 13  [2]

Figure 2: 13,598 users declared a gender in their Wikipedia profile

  2,402 users who declared themselves to be female (18%)
  11,194 users who declared themselves to be male (82%)

Sample 500 editors

edit

A sample of 500 users was then extracted from the gender-declaring Wikipedians (figure 3).[1]: 12 

 
 
 
 

Figure 3: 500 extracted Wikipedians (from the 13,598 gender-declaring Wikipedians)

  250 female Wikipedians - 50 %
  250 male Wikipedians - 50 %


How the four quartiles were formed: The breakdown of the 500 editors
The 500 editors were then divided into four quartiles based on how prolific their revisions (edits) were during their first three weeks on Wikipedia. The four quartiles were made up of editors who had made 0, 1, 2, or 4+ revisions (figure 4a).[1]: 12 

Figure 4a: Target sample of 500 new editors:
Number of edits during
his/her first three weeks
Number of editors ( % of 500 editors )
0 edits (making up 25% of the bottom 75% prolific editors)
125(25%)
1 edit (making up 25% of the bottom 75% prolific editors)
125(25%)
2 edits (making up 25% of the bottom 75% prolific editors)
125(25%)
4+ edits (making up the top 25% prolific editors)
125(25%)

Source: Antin et al [1]: 12 

There were three weeks of data unavailable for 63 editors (figure 4b).[1]: 12 

Figure 4b: Editors where the data was unavailable:
Number of edits during
his/her first three weeks
Number of editors ( % of 63 editors )
0, 1 or 2 edits (bottom 75% of prolific editors)
62(98.41%)
4+ edits (top 25% of prolific editors)
1(1.59%)

Source: Antin et al [1]: 12 

This left an adjusted number of sample editors of 437 (figure 4c).[1]: 12 

Figure 4c: Adjusted sample of 437 new editors:
Number of edits during
his/her first three weeks
Number of editors ( % of 437 editors )
0 edits
182(41.65%)
1 or 2 edits (both quartiles combined)
131(29.98%)
4+ edits
124(28.37%)

Source: Antin et al [1]: 12 

Gender break down of the top and bottom quartiles

edit
Bottom 75% of prolific editors (figures 6 and 7)
Figure 5: Editors making 0 edits in his/her first three weeks*
M/F Number ( % of 182 editors )
Editors making 0 edits in his/her* first three weeks
182(100%)

Source: Antin et al (Table 1: middle column, bottom row, by default) [1]: 13 


* As this group produces 0 edits, no 'type of revision' data is generated. Therefore, there is no gender breakdown possible for the report's tables (illustrated in figures 9, 10 and 11 below). Put more simply, if someone one makes 0 number of edits in three weeks, there is nothing to analyze. [1]: 13 

Figure 6: Editors making 1 or 2 edits in his/her first three weeks: by gender
M/F Number ( % of 131 editors )
Male editors making 1 or 2 edits in his first three weeks
66(50.38%)
Female editors making 1 or 2 edits in her first three weeks
65(49.62%)

Source: Antin et al (Table 1: middle column, bottom row) [1]: 13 


Top 25% of prolific editors (figure 7)

Note: Antin et al do not explain why the gender split in the top group of editors is not even as promised in the method on page 12, where they stated (emphasis added): "We [...] randomly sampled an equal number of men and women from each quartile to build a total sample of 500 editors."[1]: 13  The editors for whom data was unavailable only allows for one more editor (see figure 4b).

Figure 7: Editors making 4+ edits in his/her first three weeks: by gender
M/F Number ( % of 124 editors )
Male editors making 4+ edits in his first three weeks
67(54.03%)
Female editors making 4+ edits in her first three weeks
57(45.97%)

Source: Antin et al (Table 2: middle column, bottom row) [1]: 14 

Coding

edit

The report authors decided that there had to be a consensus of two-thirds of coders on a revision type (e.g. add citation) before the classification was accepted.[1]: 13 

Results

edit

Figure 8: Revisions made during the study period (total = 4,549)

  1,700 revisions made to Talk pages, User pages, and other non-article pages (37.37%)
  2,565 coded revisions (56.38%)
  284 revisions where coders were unable to reach a consensus on one or more substantive categories (6.24%)

Bottom 75% of prolific editors by gender

edit
Figure 9: Revision count by gender in the bottom 75% of sample editors
Type of
revisions
Number of revisions: female Number of revisions: male
Add citations
5(56%)
4(44%)
Add new content
26(45%)
32(55%)
Change wiki markup
28(48%)
30(52%)
Create new article
3(43%)
4(57%)
Delete content
4(40%)
6(60%)
Fix typo(s) / grammar
9(56%)
7(44%)
Reorganize existing text
3(50%)
3(50%)
Rephrase existing text
20(51%)
19(49%)
Vandalism
6(60%)
4(40%)
Unsure
3(60%)
2(40%)

Source: Antin et al (Table 1: first column) [1]: 13 


Figure 10: Number of editors making particular types of revision in the bottom 75% of sample editors
Type of
revisions
Number of editors: female Number of editors: male
Add citations
5(62%)
3(40%)
Add new content
24(46%)
28(50%)
Change wiki markup
22(51%)
21(50%)
Create new article
3(50%)
3(50%)
Delete content
4(40%)
6(60%)
Fix typo(s) / grammar
8(53%)
7(50%)
Reorganize existing text
3(50%)
3(50%)
Rephrase existing text
16(48%)
17(50%)
Vandalism
5(62%)
3(40%)
Unsure
3(60%)
2(40%)

Source: Antin et al (Table 1: middle column) [1]: 13 


Figure 11: Revision size for bottom 75% of sample editors
(using the standard metric Levenshtein distance)
Type of
revisions
Size of revision: female Size of revision: male
Add citations
190.8
1,949.5
Add new content
796.4
710.7
Change wiki markup
341.4
174.8
Create new article
2,509.7
899.2
Delete content
59.5
621.2
Fix typo(s) / grammar
16.7
254.6
Reorganize existing text
788.3
386.7
Rephrase existing text
387.2
105.9
Vandalism
987.8
288.2
Unsure
10
1.5

Source: Antin et al (Table 1: third column) [1]: 13 

Top 25% of prolific editors by gender

edit
  • The gender split for the number of revisions in this quartile was 27% female to 73% male.
Figure 12: Revision count by gender in the top 25% of sample editors
Type of
revisions
Number of revisions: female Number of revisions: male
Add citations
30(27%)
81(73%)
Add new content
162(34%)
315(66%)
Change wiki markup
231(29%)
562(71%)
Create new article
10(26%)
29(74%)
Delete content
34(37%)
58(63%)
Fix typo(s) / grammar
40(32%)
86(68%)
Reorganize existing text
9(21%)
33(79%)
Rephrase existing text
146(31%)
330(69%)
Vandalism
8(28%)
21(72%)
Unsure
20(33%)
41(67%)

Source: Antin et al (Table 2: first column) [1]: 13 


Figure 13: Number of editors making particular types of revision in the top 25% of sample editors
Type of
revisions
Number of editors: female Number of editors: male
Add citations
14(30%)
33(70%)
Add new content
43(46%)
50(50%)
Change wiki markup
48(45%)
59(60%)
Create new article
9(38%)
15(60%)
Delete content
21(47%)
24(50%)
Fix typo(s) / grammar
23(43%)
31(60%)
Reorganize existing text
9(31%)
20(70%)
Rephrase existing text
42(44%)
53(60%)
Vandalism
8(35%)
15(70%)
Unsure
15(36%)
27(60%)

Source: Antin et al (Table 2: middle column) [1]: 13 


Figure 14: Revision size for top 25% of sample editors
(using the standard metric Levenshtein distance)
Type of
revisions
Size of revision: female Size of revision: male
Add citations
542.9
230.0
Add new content
583.0
276.8
Change wiki markup
190.1
138.1
Create new article
2,019.8
1,102.5
Delete content
569.1
590.2
Fix typo(s) / grammar
489.4
175.5
Reorganize existing text
1,046.2
761.5
Rephrase existing text
265.7
133.7
Vandalism
281.4
242.2
Unsure
23.8
102.9

Source: Antin et al (Table 2: third column) [1]: 13 

Discussion and conclusion

edit
  • Male Wikipedia editors drastically outnumber female editors overall, however the report found that 18% of editors female, compared with the 2010 UNU-MERIT study which stated only 13% of editors were female.[2]
  • In the bottom three quartiles of Wikipedians, men and women made similar numbers of revisions in nearly every category of Wiki-work (fig. 9).
  • In the top 25% of Wikipedians, there is evidence that only 27% of revisions were completed by women (fig. 12).
  • In the bottom 75% of Wikipedians, women appeared to make larger revisions in many categories (fig. 11), however, only 9% of revisions were made by editors in the bottom 75%, so this may not be statistically significant.
  • In the top 25% of Wikipedians, women tended to make more sizeable revisions, particularly in the "Add New Content" and "Rephrase Existing Text" categories (fig. 14).
  • We did not see evidence that men and women are attracted to different types of editing work.
  • Two areas of work, where women made significantly larger revisions, involved creative production, synthesis, and reorganization of text, suggesting that, compared to men, women often develop more successful solutions to R&D problems posed through innovation brokers such as Innocentive.[3]
  • Despite limitations (for example, sample size, and the inability to make claims about patterns over time) this analysis suggests that the story of Wikipedia's "Gender Gap" is perhaps not as straightforward as initial reports have suggested.
  • The results are encouraging in respect that, there was significant gender parity in number of revisions, and women are better represented than the UNU-MERIT study would have led us to believe.[2]
  • Either Wikipedia has been successful in attracting more women, or there may be inaccuracies in prior studies.
  • The results are discouraging in respect that, there are far fewer women editing Wikipedia than men, and there was a particular gender skew in revision quantity among the Wikipedians who do most of the work.
  • Problematically, the most active Wikipedians are also those who largely set policies, arbitrate disputes, and do other high-level tasks into which biases of worldview and temperament can subtly creep. This is the very group among which women may need more representation, but also one which may be more difficult to break into.[1]: 13–14 

References

edit
  1. ^ a b c d e f g h i j k l m n o p q r s t u v w x y z Antin, Judd; Yee, Raymond; Cheshire, Coye; Nov, Oded (2011). "Gender differences in Wikipedia editing". Understanding Wikipedia (Conference Session). Association for Computing Machinery: 11–14. doi:10.1145/2038558.2038561. ISBN 9781450309097. Pdf.
  2. ^ a b c Glott, R.; Ghosh, R.; Schmidt, P. (2010). "Analysis of wikipedia survey, topic: Age and gender differences". UNU-MERIT. {{cite journal}}: Cite journal requires |journal= (help)
  3. ^ Bo Jeppesen, Lars; Lakhani, Karim R. (September 2010). "Marginality and problem-solving effectiveness in broadcast search". Organization Science. 21 (5). INFORMS: 1016–1033. doi:10.1287/orsc.1090.0491.