The authors have declared that no competing interests exist.
Conceived and designed the experiments: JJN DL KEK AK. Performed the experiments: JJN DL AK. Analyzed the data: JJN DL AK. Contributed reagents/materials/analysis tools: JJN YV PA. Wrote the paper: JJN DL KEK AK.
In their 2005 study, Adamic and Glance coined the memorable phrase ‘divided they blog’, referring to a trend of cyberbalkanization in the political blogosphere, with liberal and conservative blogs tending to link to other blogs with a similar political slant, and not to one another. As political discussion and activity increasingly moves online, the power of framing political discourses is shifting from mass media to social media.
Continued examination of political interactions online is critical, and we extend this line of research by examining the activities of political users within the Wikipedia community. First, we examined how users in Wikipedia choose to display their political affiliation. Next, we analyzed the patterns of cross-party interaction and community participation among those users proclaiming a political affiliation. In contrast to previous analyses of other social media, we did not find strong trends indicating a preference to interact with members of the same political party within the Wikipedia community.
Our results indicate that users who proclaim their political affiliation within the community tend to proclaim their identity as a ‘Wikipedian’ even more loudly. It seems that the shared identity of ‘being Wikipedian’ may be strong enough to triumph over other potentially divisive facets of personal identity, such as political affiliation.
Online media have become an increasingly important source of political information in recent years. This trend emerged most notably in the 2004 U.S. presidential campaign. For the first time, political blogs served as a prominent information source regarding the campaign and candidates
Much of the research examining political interaction online has provided support for a trend of polarization. One of the seminal studies in this area was Adamic and Glance's
In a similar study Hargittai, Gallo, and Kane
Political polarization has also been observed in other online contexts. For example, Feller, Kuhnert, Sprenger, and Welpe
In contrast to the findings from the blogosphere and Twitter, research on interactions in political newsgroups does not paint a clear picture of polarized interactions. Kelly, Fisher, and Smith
The aforementioned studies have provided an unclear and somewhat conflictive picture of what cross-party political interaction looks like online. In some contexts (e.g. the blogosphere, Twitter) interactions that cross ideological divides are rare. However, in other settings (e.g. online discussion boards), there is evidence for higher rates of interactions across party lines. Taken together these finding indicate that the degree of interaction and engagement with politically dissimilar others varies across contexts.
Understanding when and why people engage in political debate and discussion online is important. The degree of interaction or insularity of political groups producing political information online can have important consequences for information consumers, because it may influence the extent to which issues are presented in a biased or neutral way. The present research seeks to address this issue, and to shed light on patterns of political interaction within the Wikipedia community.
Wikipedia, currently the 6th most trafficked site in the world
Wikipedia is unique when compared to other online references. In the world of online information there is professional content, some of which aims for a neutral stance and some of which has a self-proclaimed bias, and there is user generated content (UGC), which, at its core, reflects the beliefs and ideologies of those who create it. Wikipedia is built entirely on UGC; however at the same time explicitly espouses neutrality. One of the fundamental rules of the community is that all articles must be edited from a neutral point of view (NPOV). In Wikipedia, neutrality “means carefully and critically analyzing a variety of reliable sources and then attempting to convey to the reader the information contained in them clearly and accurately. Wikipedia aims to “describe disputes, but not engage in them”
While neutrality is a fundamental principle of Wikipedia, members have a diverse array of beliefs and values. Therefore, it is particularly interesting to examine how diverse, and at times contentious, groups interact on the site. How is it that these people come together to create neutral content? Is there fragmentation, as we see in the blogosphere and on Twitter, or is there interaction and debate like we see in communities such as Usenet?
Extant research has generally failed to consider interactions among members of different political parties on Wikipedia. However, there are a couple of exceptions. One is a recent study that examined edits made to the 2012 Republican presidential candidate Mitt Romney's Wikipedia page
Social identity theory
“…self categories can be ordered in terms of a hierarchy of abstraction and include personal identities (which distinguish the person from other individuals or in-group members) and social identities (which define them as similar to other in-group members and different from out-groups on relevant dimensions). In sum, the salient self category is highly flexible and context dependent.”
Social identity can be derived from membership in a formal group (e.g. a soccer team), but can also be derived from more abstract groups or categorizations (e.g. race, gender). Tajfel and Turner
Social identity theory has been applied to the domain of politics, and research has demonstrated that people can develop social identities stemming from political party affiliation
A later study extended these findings
Membership in an online community may also be a source of social identity. Recent theoretical work by Ren, Kraut, and Kiesler
One source of common identity in online communities is task interdependence. When community members are working together to accomplish a joint task, this can foster a sense of shared identity
A study conducted by Bryant, Forte and Bruckman
It is important to note that the presence of a Wikipedian identity does not preclude the existence of other identities. Depending on the activities that a community member is performing the Wikipedian identity may or may not be activated. For example, when making edits to a particular page, the social identity a user has that is associated with that particular topic (e.g. “Republican” while editing the George W. Bush page, “soccer player” while editing the Lionel Messi page, or “mother” while editing the page on Childhood Development) may become activated.
The preceding discussion has reviewed the formation of social identity, and the influence that identity can have on intergroup dynamics. Individuals may have multiple social identities that become more or less salient depending on the social context. Members of the Wikipedia community who publicly declare their political party affiliation represent a minority of users. However, the fact that these users choose to call attention to this aspect of their identity is noteworthy. Therefore, it seems possible that either the social identity of party affiliation or of being Wikipedian could be activated in the context of this community. In the present research we examine user practices of representation and identity and examine patterns of cross-party interaction. In light of the preceding review, we pose the following research questions:
RQ1: What are the identity and representation practices of users who claim their affiliation to a party within the Wikipedia community?
RQ2: Do we see a division in patterns of participation along party lines?
RQ3: Do users exhibit a preference for interacting with members of their same political party?
RQ4: Does political affiliation of users affect the amount of conflict in discussions?
We conducted a mixed-methods analysis of patterns of activity, interaction, and identity representation practices among 1,390 members of the Wikipedia community who explicitly proclaimed their political affiliation as either a Republican or Democrat. In order to determine user political affiliation, user pages were examined. Content analysis was used to evaluate user representation practices in user profiles, and to categorize and thematically group the most edited articles by members of each party. Social network analysis was used to explore the research questions regarding patterns of interactions. Finally, content analysis was used to examine differences to conflict across and within parties, and to evaluate user representation practices in user profiles.
On personal user pages users have the ability to display userboxes. A userbox is “a small colored box designed to appear only on a Wikipedian's user page as a communicative notice about the user, in order to directly (or indirectly) help Wikipedians collaborate more effectively on articles”
Members of the Wikipedia community have the option of creating a customized user page. Pages can be personalized to reflect the preferences and interests of the individual users, and one of the primary ways that users personalize their pages is through the use of userboxes, which were described above. A qualitative analysis of the userboxes of a randomly selected sample of fifty Democratic and fifty Republican users was conducted.
First, the overall number of userboxes for each user was tallied. Next, the number of political party boxes that a user listed on his or her page was tallied. A box was counted as a party box if it explicitly expressed support for, or membership in, the Republican or Democratic Party. The number of politically oriented userboxes that were
Content analysis of user walls was performed on a randomly selected subsample of 100 user pages (50 Republicans, 50 Democrats). Intercoder reliability was assessed using Holsti's reliability score
To verify that these subsamples represent the original samples, t-tests have been performed comparing the registration time and the logarithm of the number of comments and edits plus one The logarithm was used to account for the heavy tailed nature of these distributions. Details on the corresponding distributions are given in the following section. In all cases and for both groups of partisan users, the null hypothesis that sample and subsample are extracted from the same distributions cannot be rejected.
Activity and interaction data came from a complete dump of the English Wikipedia, dated March 12th 2010. First, we considered edit activity. We counted all edits made by users in our sample to each Wikipedia article.
Inset shows the same data in log-log scale. In the case of all users only users with at least one edit are considered.
We also extracted comments written by users in talk pages, i.e. special wiki pages devoted to communication among editors (also included in the above mentioned complete dump). We considered both
The distribution of the number of comments per user in article talk pages (
Inset shows the same data in log-log scale. In the case of all users, only users with at least one comment are considered.
A further comparison between the partisan users and the overall Wikipedia users is made in
Note that the depicted values are per month for the set of all registered users (black curve with diamonds) while for the two groups of partisan users values are averaged over two month periods to filter out fluctuations due to the smaller amount of users in these groups.
To analyze patterns of communication among our set of users, we identified two networks of interactions based on messages written by the users in talk pages. From article talk pages we extracted a
Nodes | Dem | Rep | Edges | Giant comp. | Reciprocity | Clustering coefficient | Average distance | ||
Reply | 270 | 161 | 109 | 430 (854) | 83.9% | 0.28 | 0.040 | 4.75 | |
Wall | 434 | 258 | 176 | 997 (2619) | 95.6% | 0.29 | 0.074 | 4.00 |
Basic statistics of the two networks based on replies in article talk pages (
Lists of the most frequently edited articles among Democrats, Republicans, and all Wikipedia editors were generated. The number of articles shared by various groups was calculated. The articles were also coded according to topics. Articles were coded as “Political” (relating to a political issue or a politician, e.g. United States Presidential Election, 2008; George Bush) or “Not Political” (e.g. Britney Spears, 2008 Summer Olympics). Political pages were further coded as “Conservative” (related to a conservative politician, commentator, or issue, e.g. Rush Limbaugh), “Liberal” (related to a liberal politician, commentator, or issue, e.g. Al Gore), or “Neutral” (political in nature, but not partisan, e.g. European Union, September 11 attacks).
To assess whether there is a preference for interaction among editors belonging to the same party, or to different parties, we studied the mixing coefficient of the networks, and we performed a shuffle test in order to assess statistical significance.
We first extracted from each network a matrix representing how many connections (based on comments) are directed from a Democrat to a Democrat, from a Democrat to a Republican, and so forth. We then normalized these matrices and computed the mixing coefficient as the preference for inter-party or for intra-party interaction, according to Newman
We examined discussion thread conflict in order to assess whether or not users exhibit different interaction styles with same party members versus different party members. All discussion threads on article talk pages that included at least two Democrats and no Republicans (572), at least two Republicans and no Democrats (147), or at least one Democrat and one Republican (583) from our sample were extracted from the data set. From this sample, we then selected all threads related to articles that dealt with political or other potentially controversial topics. Examples include “War on Terrorism”, “Mike Huckabee”, “Eliot Spitzer”, and “Mahmoud Ahmadinejad”. These threads were then coded for whether or not they were conflictual, and for whether or not the conflict was political in nature. This resulted in 144 threads with at least one Democrat and one Republican, 130 threads with two or more Democrats, and 71 threads with two or more Republicans. Holsti's reliability score
First, we tested to see if there were differences in the average number of userboxes listed on the profiles of Republicans and Democrats. There was no significant difference (p = 0.3). Unsurprisingly, Republicans (M = 3.06, SD = 5.37) had a significantly higher number of conservatively valenced user boxes than Democrats (M = 0.08, SD = 0.44) (
Total n° of Boxes | Political Boxes | |||
Conservative | Liberal | Other | ||
Democrats | 49.24 (±5.51) | 0.08 (±0.06) | 2.51 (±0.49) | 2.20 (±0.46) |
Republicans | 42.90 (±7.56) | 3.06 (±0.76) | 0.27 (±0.09) | 2.52 (±0.76) |
Diff. significant? | No | No |
Values in parenthesis indicate the corresponding standard error of the means. Bottom row indicates the outcome of a t-test (n = 50) for a significant difference between the mean values of the supporters of the two parties.
Party Boxes | Wikipedia Boxes | difference significant? | |
Democrats | 1.44 (±0.19) | 4.70 (±0.52) | p<0.01 |
Republicans | 1.26 (±0.13) | 3.16 (±0.56) | p<0.001 |
Values in parenthesis indicate the corresponding standard error of the means. Last column indicates the outcome of a paired t-test (n = 50) for significant difference between the mean values of party and Wikipedia boxes within the supporters of the two parties.
Out of the 100 most edited articles, Democrats and Republicans had 44 articles in common. For Democrats, 38 of the top 100 most edited articles dealt with political topics. Out of those, 15 were coded as liberal, 15 were coded as conservative, and 8 were coded as neutral. Thirty-five out of the top 100 most edited articles by Republicans dealt with political topics. Of these, 7 were coded as liberal, 17 were coded as conservative, and 11 were coded as neutral. These findings stand in contrast to the most edited articles for users in general, only 22 of which dealt with political topics. Of those, 3 were coded as conservative, 5 were coded as liberal and 14 were coded as neutral.
Rank | Democrats | Republicans | All Users |
1. |
|
|
|
2. |
|
|
Wikipedia |
3. |
|
United States | United States |
4. | Unites States |
|
|
5. |
|
|
Adolf Hitler |
6. |
|
|
Michael Jackson |
7. | Wikipedia | Wikipedia | Britney Spears |
8. | Britney Spears |
|
Jesus |
9. |
|
Virginia Tech Massacre | World War II |
10. |
|
Adolf Hitler | PlayStation 3 |
Democrats | Republicans | |||
Democrats | 193 | (384) | 94 | (215) |
Republicans | 86 | (180) | 57 | (75) |
Democrats | Republicans | |||
Democrats | 395 | (1263) | 243 | (665) |
Republicans | 187 | (519) | 172 | (738) |
The results of a shuffle test for assortativity, shown in
Blue nodes represent Democrats, and red nodes Republicans. The size of each node is proportional to the number of connections (degree) in the unweighted reply network. Edges connecting two Democrats are depicted in blue, edges connecting two Republicans in red, edges connecting a Democrat and a Republican in green.
r | rrand (avg) | σ |
Z-score | |
Reply (unweighted) | 0.070 | 0.0028 | 0.0505 | 1.33 |
Reply (weighted) | −0.062 | −0.0122 | 0.0316 | −1.59 |
Wall (unweighted) | 0.095 | −0.0053 | 0.0301 |
|
Wall (weighted) | 0.237 | −0.0017 | 0.0131 |
|
r represents the mixing coefficient in the real network; rrand (avg) and σ
An examination of the interactions on user walls shows a different pattern, indicating a significant preference for interaction among members of the same party (Z-score = 3.33 for the unweighted network and 18.22 for the weighted network, see
Finally, we examined levels of conflict in discussion threads that dealt with political or other potentially controversial topics. There were relatively high levels of conflict across all three groups of threads that we examined. This is not a particularly surprising finding, given that we purposefully selected threads that dealt with controversial topics. Seventy-four percent of the threads that involved at least one Democrat and one Republican, 65% of threads involving at least two Democrats, and 77% of threads involving at least two Republicans were conflictive. See
Involving a Democrat and a Republican | Involving two Democrats | Involving two Republicans | |
Total number of threads | 583 | 572 | 147 |
Political threads | 144 | 130 | 71 |
Political threads displaying conflict | 106 | 80 | 53 |
Our results paint an interesting and somewhat mixed picture of the nature of interactions among members of the Wikipedia community who espouse a political affiliation. First, we examined identity representation practices. We found that a subset of users on Wikipedia publicly proclaim their political affiliation through userboxes, and users who proclaim their affiliation for a particular party tend to have high numbers of userboxes that are ideologically aligned with that party. However, these ‘political’ users also had equally high numbers of Wikipedia userboxes. That is, boxes that espoused an identity of being a ‘Wikipedian’. The results indicate that the social identities of being a member of a political party and of being a Wikipedian may be equally important. Analysis of patterns of activity and interaction indicates that which identity is activated may depend both on user context and the nature of activities in which users are engaged.
An examination of the most edited articles for each group reveals that Democrats and Republicans both exhibit a tendency for editing articles that deal with political topics. For both groups, roughly one-third of the most edited articles dealt with political topics, compared to less than one-quarter for users in general. Interestingly, for both groups we find a preference for topics directly related to their party, such as “Barack Obama” or “Bill Clinton” for Democrats, and “John McCain” or “Republican Party” for Republicans.
Laniado, Castillo, Kaltenbrunner and Fuster-Morell
However, we did see evidence for preference to interact with members of the same party in user walls. It is interesting that we observe this tendency in these more personal spaces, but not on article talk pages. It may be that in the course of conducting activities that are central to the Wikipedia community (e.g. editing articles), the identity of being a Wikipedian is activated and, as a result, the political identity is not salient. In the context of interactions on user walls, where personal activities take greater precedence, the importance of political ideology may shine through more strongly.
Finally, we found that levels of conflict were high both within and across parties when the discussion threads dealt with political or other potentially controversial topics. Interestingly, there were a significantly greater number of conflictive cross-party and Republican threads, indicating that Democrats have lower rates of within party conflict in the context of these controversial threads.
Although Democrats and Republicans seem to maintain their political identity within the Wikipedia community, our findings show that users displayed more “Wikipedia” boxes than political boxes on their user pages, indicating that the identity of being a Wikipedian may be more salient in the context of this community. Further, the lack of preference to interact with same-party members in the context of article discussions does not indicate the same polarization that has been observed in other contexts