Bridges or Hate-Links? The Meaning of Common Links in Media Conflict

When groups have strong disagreements online, we tend to worry about echo chambers — the possibility that each group is having conversations that are isolated from each other. If these echo chambers persist, might an agreement ever be reached? These worries extend to echo chambers in personalization algorithms (Sunstein 2014), echo chambers in social news sharing (Pariser 2011; Bakshy, Messing, Adamic 2015), and echo chambers in the link structure of blogs and news (Drezner and Farrel 2004, Adamic and Glance 2004).

For example, Gilad Lotan’s analysis of media coverage during the 2014 Israel-Gaza War showed that on social media, pro-Israeli and pro-Gaza groups were following very different information sources, with the exception of the Haaretz news organization. Building on that network analysis, Lotan pointed out that Haaretz formed an important bridge between these two groups in conflict.

When we see these polarized network maps, it’s easy to think that the bridges between them offer some hope of escaping these echo chambers. But opposing sides don’t always follow or link to something for the same reason. Sometimes, opposing sides link to material to disagree with it or make straw-man arguments (Hargittai 2008). At such times, overlapping links can draw out the core disputes in a controversy, with what Zeynep Tufekci calls “hate-linking” (2014). To show how this happens, we examine how opposing sides linked to information sources in the Gamergate controversy.

(This blog post was one of Alyssa’s projects during her time as a summer intern at the Berkman Center. We are especially grateful the generous data collection support offered by David Larochelle, Whitney Erin Boesel, and Hal Roberts.)

What is Gamergate?  

On August 16 2014, game designer Zoe Quinn’s ex-boyfriend Eron Gjoni posted a lengthy, very public blog post about their relationship intended to portray Quinn as dishonest and untrustworthy.  Some people read Gjoni’s post and drew the conclusion that Quinn had been leveraging personal relationships for positive reviews.  Gjoni offered no evidence that any conflict of interest of that nature existed at the time. Nonetheless, gamers who had been mobilizing on the Internet very quickly began discussing and investigating what they saw as a scandal in game journalism; according to the movement that we now know as GamerGate, nepotism and dubious ethics have been long-standing issues in game journalism.

Gamergate most visibly manifested itself as a harassment campaign first against Quinn and later against Phil Fish (who supported Quinn); Anita Sarkeesian (a feminist game critic); and Brianna Wu (a game developer). This harassment campaign involved death threats, rape threats, doxing, swatting, and a mass shooting threat that led Sarkeesian to cancel a talk at Utah State University. Gamergate participants claim that the majority of the movement was not involved in the harassment; members point to changes in publications’ ethical policies and ads pulled from websites critical of Gamergate as evidence of Gamergate’s impact.  They claim to have raised awareness about issues in gaming through the hashtag campaigns (in which women and minorities, some of whom were fake accounts, tweeted in support of Gamergate) and leaked emails from gaming mailing lists (which they argued was evidence of collusion amongst gaming journalists).  Commentators have suggested that the Gamergate controversy is evidence of upheaval in the gaming community, both over the role of journalists and the increasing diversity in the community.

What Gamergate Events Were Covered By the News?

We can track how many sentences were relevant to the Gamergate controversy over a period of several months using MediaCloud’s Dashboard search feature. MediaCloud monitors every article published by more than 50 thousand media sources every day. We wrote a query to find articles relevant to Gamergate from the beginning of August 2014 until April 2015. In this case, we also downloaded everything that those articles linked to. The below graph plots the number of sentences about Gamergate per day across all these sources.

One of the earliest peaks in the graph corresponds to a series of “Gamers are Dead” articles published around August 28, 2014.  These articles were a reaction to the early actions (i.e. online harassment and doxxing) by the nascent Gamergate movement.  Many authors said that the “gamer” identity was becoming irrelevant as the gaming becomes increasingly diverse. These articles contextualized Gamergate as a backlash to that diversification by a demographic (young white men) who are used to being central to gaming communities and identities.  The Gamergate movement took the near-simultaneous publication of so many articles with the same point of view as evidence of anti-Gamergate collusion by journalists.  

Some of the peaks in the graph can be traced back to major Gamergate-related events that were prominently featured in mainstream news. This includes Anita Sarkeesian cancelling her talk at Utah State after death threats and the Gamergate-inspired Law and Order episode that aired in February, briefly reawakening the dialogue.  

Several peaks didn’t seem to have a precipitating event. To investigate those cases, we generated a list of the news articles matching the Gamergate query and looked for unifying factors. Some of these peaks seem more associated with controversies in Wikipedia editing. After looking at trends in wikipedia edits and the articles within those peaks, it seems like they corresponded with spikes in Wikipedia editing, but it’s hard to know which influenced which.

In September 2014 and in early 2015, the Wikipedia article on the Gamergate controversy became a battleground for control of public awareness about the controversy. In September, Wikipedia’s founder, Jimmy Wales, weighed in on Twitter, and eventually Wikipedia’s Arbitration Committee agreed to arbitrate the conflict.  ArbCom banned or sanctioned editors from Gamergate as well as editors who had tried to undo repeated pro-Gamergate vandalism, sparking a dialogue about Wikipedia’s internal politics and its gender disparities.  

Is There Common Ground Across Pro and Anti Gamergate?

Given the polarizing nature of the Gamergate discussion, it’s helpful to ask what, if any common ground the two sides of the controversy had — it is difficult to believe that people on either side made sense of the dispute in completely different ways. Perhaps the interests shared by both sides might illuminate commonalities in their worldviews.

Common ground doesn’t always mean agreement. When one side links to another’s dialogue, it may not be to engage meaningfully with the other side’s opinions, but rather to discount their arguments or inform their audience of their opponents’ actions. We can also ask a parallel question: what parts of one side’s dialogue was the other failing to listen to?  Were pro-Gamergaters not talking about things anti-Gamergaters cared about and vice versa?

Several others have carried out network analysis of the #GamerGate hashtag on Twitter, including Andy Baio’s 72 hours of #GamerGate on Twitter and Brian Keegan’s followup analysis of Baio’s dataset. Csefalvay’s working paper on ecological measures of diversity uses a #Gamergate dataset, and the ggautoblocker system deployed during the controversy also uses Twitter networks. Michael Trice’s fascinating research on documentation practices of #GamerGate also contextualizes Twitter usage in relation to other document-sharing approaches used within the movement. Finally, Roopika Risam’s article on hashtag feminismdiscusses intersections with GamerGate, and Zangerle, Schmidhammer, and Specht describe Gamergate-related tweets about Wikipedia.

In this post, we look at a wider network of blogs and news articles that also includes tweets.  

Networks of Link Sharing in the GamerGate Controversy

In the figure below, each node (colored dot) represents a news article that is part of the MediaCloud Gamergate controversy.  Within the MediaCloud dataset of 11,368 stories about Gamergate, we chose the 50 most highly-ranked (in-linked), clearly partisan articles (coded manually) as starting points to explore the main Gamergate dialogue.  The blue articles on the left are anti-Gamergate articles, the red articles on the right are very strongly pro-Gamergate articles, and the pink articles on the right are more moderate pro-Gamergate articles. We distinguish between moderate and strongly pro-Gamergate articles because a clear difference in outlook is evident between them.

Next, we explored the articles our starting points linked to, and the articles those articles linked to and so on until we had seen all articles within five degrees of separation from our starting points, calculating the “degree” between highly partisan nodes and other articles. The links in this graph represent these composite ties; the stronger the link between two articles, the fewer degrees of hyperlink separation between them. The final graph contains 477 articles; a small fraction of the 11368 stories in the MediaCloud results. From close manual examination of the articles involved, we are confident that they are a good representation of the discussions at the heart of the controversy.

To understand what discussions were of high importance within the larger Gamergate controversy, we manually categorized the top 100 most inlinked articles in the subset of 477. Topics included articles about online harassment (orange), “gamers are dead” articles pointing out diversity among gaming audiences (yellow), articles about games journalism (cyan), and primary source material (green). These articles are represented by smaller circles.

In the network graph shown above, the closer two nodes (corresponding to articles) are, the fewer degrees of hyperlink separation between them.  A node’s position from right to left roughly corresponds to their link-degree on the pro- or anti- Gamergate side of the controversy.  In this way, we can illustrate how often and closely the source articles link to the same material — and how often or closely they ignore material that the other side of the controversy is linking to.

In this illustration of the 100 most-inlinked articles, roughly four in five (~80.1%) articles are only linked to by one side of the controversy. Approximately 20% of articles depicted in the graph are bridges, linked to from both sides of the controversy.

What Were the Echo Chambers in the GamerGate Controversy?

Many of the most highly inlinked anti-Gamergate articles discuss Gamergate in terms of the harassment its members have carried out. In articles mostly ignored by Gamergate, they are described as “dishonest fascists,” “trolls” and members of a “hate group” in article headlines.  

Articles shared primarily by Anti-Gamergate sources also tended to dispute the idea that Gamergate was actually resolving its ethical issues. Others dispute that ethics were as pressing as the movement claimed them to be. These articles emphasize the vitriolic harassment perpetrated against those (primarily women) who have spoken out against Gamergate and suggest that the Gamergate backlash is a symptom of a larger cultural reaction to diversification of previously homogenous spaces, of which video gaming is only one.  

The pro-Gamergate dialogue mostly discusses the controversy in terms of ethical violations and allegations against Zoe Quinn; many of the headlines of top-ranked articles focus on discounting Quinn’s credibility– other articles discuss allegations of corruption and collusion within gaming journalism.  

Bridges as Points of Conflict Rather Than Common Ground?

Other analyses of network bridges, like Gilad Lotan’s analysis of the 2014 Israel/Gaza conflict, interpret link bridges as an important site of common ground. In the Gamergate controversy, the two sides often linked to common sources for very different reasons.

BRIDGE 1: STORIES ABOUT ETHICS IN GAMES JOURNALISM

In the network graph, 75% of articles manually labeled as ethics and ad-related received links from both sides. Many of the articles shared by both sides were about the success of campaigns by Gamergate to get advertisers to pull online ads from publications whose actions and/or policies they deemed suspect. Both sides also linked to publications’ statements clarifying their ethics policies.  Gamergate sources shared these articles as evidence of success — their social media and email campaigns were effecting change in the gaming journalism space, and their voices were being heard by powerful corporations like Intel.  Anti-Gamergate sources agreed that Gamergate was being heard, but they contextualized Intel’s ad-pulling as the success of a persistent bullying campaign.  They also linked much more than Gamergate did to an article by Leigh Alexander questioning whether Gamergate’s stated concerns about ethics were the most pressing issues facing the gaming industry. Gamergate also linked with less degrees of separation than anti-Gamergate to Ars Technica’s statement regarding allegations of collusion and a blog post about allegations of collusion among gaming journalists ruining careers.  Each side attempted to put their own spin on major events that garnered press coverage and formal statement by journalists (i.e. ad-pulling) while not engaging with the issues that were claimed as major concerns by the other side.  

BRIDGE 2: STORIES ABOUT CHANGES IN GAMER CULTURE

Another important theme of media coverage during GamerGate focused on changes in gaming demographics and culture.  Both sides link to these “gamers are dead” articles. Gamergate points to them evidence of gaming journalists’ anti-gamer collusion, but anti-Gamergate groups disagreed. Instead, anti-Gamergate articles tended to frame those articles in terms of the diversification of gaming and the need for the predominately white male self-identified “gamer” demographic to make room for change to happen. Gamergate participants described the same articles as an attack on a gamer identity that had become a casualty of overly touchy “political correctness.”

One of the foundational arguments Gamergate uses to support allegations of conspiracy amongst gaming journalists is the cluster of “Gamers are Dead” articles that appeared on and around August 28th, 2014.  This supposed threat to the “gamer identity” became a strong rallying cry for Gamergate to gather support, while the timing of the articles gave traction to accusations of journalists conspiring to advance a progressive agenda.  Anti-Gamergate sources, in contrast, expressed excitement that gaming is becoming inclusive and that marginalized people have many more opportunities to participate and create.  

Both sides link to the “Gamers are Dead” articles with approximately the same prevalence, however, although the cluster of “Gamers are Dead” articles is slightly more strongly linked to by anti-Gamergate sources.  These articles represent common context — both sides are talking about them and their implications — but not common ground.  While pro-Gamergate describe them as an attack on what was previously a cohesive identity, anti-Gamergate cites them as evidence of cultural evolution — certainly upheaval, but in a progressive direction.  

BRIDGE 3: STORIES ABOUT ONLINE HARASSMENT AND GAMERGATE

Although some articles about online harassment exclusively received links from one side, other articles about online harassment are linked equally by pro-gamergate and anti-gamergate sources.

Articles written from a strongly pro-Gamergate perspective claim harassment victims are responsible for the harassment they faced and dispute the severity of death threats sent to Zoe Quinn and Anita Sarkeesian, among others. Some argue that feminist commentators and developers are leveraging threats for attention on the Internet. Other articles play to both sides, claiming that both sides are engaging in harassment and that the problem of harassment against women is beyond solving. Articles written from an anti-Gamergate perspective tend to strongly criticize the pro-Gamergate harassers as anti-feminist and argue that their grievances are groundless.  When anti-Gamergate sources link to these articles, they take approximately the same viewpoint as the authors on the problem of harassment. On the other hand, pro-Gamergate sources either dispute the commentary when linking to an articles or engage in hand-wringing over the problem of harassment.  

Rethinking the Meaning of Common Links during Media Conflict

In a conflict like Gamergate, sources that both sides of a dialogue have in common do not necessarily represent opinions both sides share, but they do represent things both sides care about. Even when they link to the same articles, each side may frame those links in very different ways. Links to articles about ethics in games journalism celebrated or lamented Gamergate’s early success. Links to stories about changes in gamer culture lamented or celebrated those changes. Finally, articles about online harassment received links for a variety of reasons, even within a given side.

So-called “hatelinking” — linking to articles with opinions or events that don’t promote one’s cause — may give networks the appearance of having conversations of common ground, but analysis of context and commentary reveals that these connections don’t necessarily represent meaningful cross-talk.

Even if the apparent bridges between networks in conflict can just be further evidence of conflict, we shouldn’t be too hasty to give up on the possibility of shared understanding. Even the most vigorous attempt to debunk a source shows that you take it seriously. If a hate link is the only way that someone can learn what the other side thinks, perhaps those connections are even more precious.