This may be the author’s version of a work that was submitted/accepted for publication in the following source:

Burgess, Jean& Matamoros Fernandez, Ariadna (2016) Mapping sociocultural controversies across digital media platforms: one week of #gamergate on , YouTube, and . Communication Research and Practice, 2(1), pp. 79-96.

This file was downloaded from: https://eprints.qut.edu.au/95337/

c Consult author(s) regarding copyright matters

This work is covered by copyright. Unless the document is being made available under a Creative Commons Licence, you must assume that re-use is limited to personal use and that permission from the copyright owner must be obtained for all other uses. If the docu- ment is available under a Creative Commons License (or other specified license) then refer to the Licence for details of permitted re-use. It is a condition of access that users recog- nise and abide by the legal requirements associated with these rights. If you believe that this work infringes copyright please provide details by email to [email protected]

Notice: Please note that this document may not be the Version of Record (i.e. published version) of the work. Author manuscript versions (as Sub- mitted for peer review or as Accepted for publication after peer review) can be identified by an absence of publisher branding and/or typeset appear- ance. If there is any doubt, please refer to the published source. https://doi.org/10.1080/22041451.2016.1155338 Mapping sociocultural controversies across digital media platforms: One week of #gamergate on Twitter, YouTube and Tumblr Forthcoming in a special issue on Digital Methods of the journal Communication, Research & Practice. Preprint version, please cite the final version when published.

Jean Burgess (corresponding author)

Ariadna Matamoros-Fernández

QUT Digital Media Research Centre, Queensland University of Technology, Brisbane, Australia

QUT Digital Media Research Centre, Z1-515, Creative Industries Precinct, Queensland University of Technology, Victoria Park Road, Kelvin Grove, QLD 4059, Australia. Phone: +61 7 3138 8253

Dr Jean Burgess (@jeanburgess) is Professor of Digital Media and Director of the Digital Media Research Centre (DMRC) at Queensland University of Technology. Her research focuses on the cultures, politics, and methods for studying social and mobile media platforms.

Ariadna Matamoros-Fernández (@andairamf) is a PhD candidate at Queensland University of Technology and member of the Digital Media Research Centre (DMRC) with a background in . Her research seeks to understand the cultural dynamics of race and identity in Australian social media. Mapping sociocultural controversies across digital media platforms: One week of #gamergate on Twitter, YouTube and Tumblr

Social media play a prominent role in mediating issues of public concern, not only providing the stage on which public debates play out, but also shaping their topics and dynamics. Building on and extending existing approaches to both issue mapping and social media analysis, this paper explores ways of accounting for popular media practices and the special case of ‘born digital’ sociocultural controversies. We present a case study of the with a particular focus on a spike in activity associated with a 2015 Law and Order: SVU episode about gender-based violence and harassment in games culture that was widely interpreted as being based on events associated with Gamergate. The case highlights the importance and methodological challenges of more adequately accounting for the cultural dynamics of digital media within and across platforms.

Keywords: social media; digital culture; games culture; gender; issue mapping; Twitter

Introduction

If you haven’t heard of #GamerGate, lucky you. If you have, and you have an opinion about it, you probably fall into one of two camps. You’re in the camp that thinks it’s a Web-based movement of upset about a perceived lack of ethics among video games journalists. Or you’re in the camp that thinks it’s a Web-based campaign of harassment against women who make, write about and enjoy video games, masquerading as a movement of gamers upset about a perceived lack of ethics among games journalists. (Wofford, 2014).

In this paper we outline a novel approach to applying controversy analysis and issue mapping to social media, and to demonstrate how this approach can be used to account for popular media practices and the special case of ‘born digital’ sociocultural controversies—that is, controversies that originate and are substantially enacted via digital media. We do so through a case study of GamerGate (GG), with the #gamergate

Twitter used as an initial anchoring point for the cross-platform exploration of the controversy. We confine our analysis to a particular week in the unfolding history of the social media activity around the keyword ‘gamergate’, during which a Law &

Order: SVU episode directly based on gender-based harassment and violence in videogame culture, and explicitly ‘ripped from the headlines’ around the GamerGate controversy, was broadcast. We further refine our focus to concentrate on the key media objects circulating during that period. As we outline the method and describe our findings, we walk through the three principal steps of issue mapping as applied to social media:

• Building an issue inventory;

• Mapping the issue networks; and

• Identifying the key mediators.

We suggest a number of additional metrics and analytics that would be needed in order to properly account for the role of other digital media platforms; and the further supplementary qualitative analyses that are needed to make sense of cultural aspects of sociocultural controversies - aspects like intertextuality, humour, and visual culture. The paper concludes with an outline for future issue mapping studies that may more adequately account for the medium-specificity of social media controversies, rather than treating social media platforms as transparent sources of behavioural data on social issues; as well as a discussion of the limitations of publicly available data, especially like #gamergate which are oriented towards particular perspectives on controversial issues.

Theoretical framework

This paper is concerned with the role of digital media in giving rise to and coordinating communication within issue publics. We understand ‘publics’ to be emergent socio- political assemblages with shared or interlocking concerns who know themselves as, and act as, publics through media and communication. Digital media platforms and practices influence both the nature of publics and the means through which they engage in issues (Bruns & Burgess, 2015). The internet, and especially social media, have troubled any clear division between ‘public’ and ‘private’ topics, discourses and spaces

(Papacharissi, 2010; 2015); and some have argued that it is a constitutive feature of the digital media environment that publics are increasingly ‘issue-ified’ (Marres, 2015). We note here that there is no necessary correspondence between the public’s degree of visibility (in the vernacular, its ‘publicness’ as opposed to its privacy), and publics as formations of parties (who may not be ‘publicly’ visible) to an issue. Post-Habermasian critiques and interventions over the past several decades (e.g. Fraser, 1990) have made it clear that publics can mobilise in private; that they need not be explicitly co-present, and—crucially for the issue mapping method—may be called into being by the mere implication that others are engaging with the same media, cultural or social experiences as we are, rather than through direct discursive interactions (Berlant, 1997; Livingstone,

2005; Warner, 2002). While these theoretical debates are not within the scope of this paper itself, they do produce significant caveats for our findings and further methodological challenges, which we return to in the limitations and conclusions sections.

We define issues as: a. matters of shared concern; that b. involve uncertainty and/or disagreement—but that are not necessarily binary debates, and can be multi- sided. Such issues range from environmental contestation around land use to everyday experiences of gender and race. Many such issues impact directly on our everyday lives; we share and consume news stories about them and personal media representations of our engagements with them via our everyday social media use, as well as explicitly discussing or deliberating upon them. Issue publics therefore leave behind plentiful traces as rich, multimedia social media data, opening up powerful new opportunities for digital media research, and presenting significant challenges both to theories of publics and to empirical orthodoxies for studying them. Among these challenges is the need not only to take into account the material politics of the platforms and devices via which all this occurs (Gillespie, 2014; Marres, 2015; Massanari, 2015); but also—as we demonstrate in this paper—to understand the role of social media objects (photographs, videos, hashtags) as key mediators in issue publics.

Issue publics are animated by acute controversies which are of a different ontological order to issues: they are discrete and identifiable sites of uncertainty and (as well as disagreement) around a given issue (Callon, Lascoumes, & Barth,

2001). In the digital media environment controversies are given an extra boost of energy by algorithmic curation models (Gillespie, 2014) like ‘trending topics’ that amplify spikes in activity. Here we align ourselves with the traditions of Science and

Technology Studies (STS) research within which controversies provide a crucial analytical focus, not only because studying controversies helps to understand the identities and arguments of people involved in a particular issue, but also because they are generative, bringing new actors onto the field (both human and non-human), connecting existing issues to each other in new combinations, and propelling issues forward (Callon, Lascoumes, & Barth, 2001). An empirical focus on controversies as generative ‘acute events’ (Burgess & Crawford, 2011) in the life of issues is therefore fundamental to understanding how both publics and issues emerge, engage and overlap, and we incorporate some of the principles and theoretical frameworks of controversy analysis into our approach. Applying issue mapping to social media

Since the early 2000s, web-based methods have been used for issue mapping—a set of approaches that use web data and network analysis to trace the relations among stakeholders, arguments, and objects in uncertain topics—especially scientific and technical topics (like climate change, genetically modified food, and energy policy) involving both lay-people and clearly delineated experts (Rogers & Marres, 2000;

Rogers, Sánchez-Querubín, & Kil, 2015; Venturini, 2012). Issue mapping using digital methods and informed by controversy analysis therefore provides a highly focused methodology for the study of how issue publics emerge through their engagement with specific controversies related to matters of shared or interlocking concern (Marres,

2015).

However, issue mapping methods are yet to be adequately adapted for the study of the publics that emerge and are sustained via social media. The current deficit has two aspects: the narrowness of its primary focus on science and technology topics; and the consequent blindness to the medium specificity of social media as opposed to the internet in general (although there have been some very recent developments in relation to this second problem – Marres, 2015; Marres & Moats, 2015). Currently the approach tends to focus mostly on expert-oriented topics like science communication, is oriented toward the formal spaces and institutions of democracy, and until very recently has neglected the shaping affects of the social media platforms and algorithms that mediate so much contemporary public activity (Marres, 2015). Social media’s convergence between ‘public’ and ‘private’ cultures requires that controversies related to sociocultural issues like, for example, the gender politics of online communication

(addressed in the present paper) be taken seriously. A focus on image-sharing, memes and emotion in social media is also needed, because popular and everyday modes of communication (from selfies to memes) are fundamental to the cultural dynamics of social media platforms and play a significant role in the ways that certain kinds of controversies emerge, evolve and impact on matters of public concern.

The present paper, therefore, emerges from a body of work that seeks to develop digital methods adequate for the study of issue publics in social media. Our research design, while necessarily customized for each particular issue or controversy, broadly adheres to the following sequence of activities:

(1) Broad issues that may give rise to acute controversies are monitored on an

ongoing basis by tracking bundles of keywords and hashtags associated with

them using a local installation of the DMI-TCAT tools for Twitter data capture

and analysis. Ongoing monitoring of the TCATs is used to identify spikes in

activity that indicate acute controversies in the form of social media events. For

each issue a dossier of background and ongoing media research is compiled

using shared online notebooks.

(2) Once identified as productive for further study, the controversy’s antecedents

and genealogy in both expert discourse and popular culture are traced using

further secondary research and manual review of key media resources that have

emerged from the data as salient.

(3) Following the model established in pilot analyses of the #agchatoz issue network

(Burgess, Galloway, & Sauter, 2015), a basic inventory that can be used for

further analysis is developed from this first batch of Twitter data: describing and

quantifying over time the levels of activity, key actors and their relationships,

hashtags and URLs associated with the controversy as represented by Twitter

activity. (4) Gephi is used to analyse, visualize and further explore a variety of

associations—not only among Twitter accounts but among media objects such

as hashtags and videos. Techniques such as Associational Profiling (Marres,

2015) which layers themes (using hashtags and other keyword tags as proxies)

over network clusters (using both accounts and media objects as nodes) can be

used to develop a sense of how these clusters represent different perspectives

and experiences on the controversy and its related issues.

(5) Qualitative analysis of the themes and media objects associated with issue

network clusters and sub-controversies is conducted using temporally

representative sampling, open coding and grounded theory. The combination of

dominant platforms, forms and genres that constitute the issue network for each

controversy is highlighted (e.g. an emphasis on news sources and Twitter

debates vs an emphasis on YouTube videos and ‘meme’ images).

The present paper is particularly novel because it further adapts these methods to the study of acute controversies that engage with the politics of social media platforms themselves. The paper describes and reflects on our experimental application of this approach to digital culture through the analysis of one week of social media activity concerning the #gamergate controversy. This controversy brings together the politics of digital media and videogame culture, gender politics and online harassment; and because of its embeddedness in digital culture, necessitates both an ontological and political emphasis on objects associated with digital media’s ‘platform vernaculars’

(Gibbs et al, 2015)– images, videos, memes and hashtags. The #gamergate case study and Intimidation Game sub-controversy

Issue background

In order to explore the current possibilities and future challenges of this approach, we selected a long-running, high-intensity controversy: GamerGate. Arguably by design, the GamerGate ‘movement’ is difficult to define, and its history is torturously complex; one of the reasons it is a good topic for issue mapping is the multiple competing frames and definitions of the issues in question – pro GamerGaters insist that it is about ‘ethics in gaming journalism’; progressives and feminists locate it as a reaction to the diversification of gaming culture (Todd, 2015) and focus on the sexual harassment, violence and abuse directed towards them by anti-feminist members of the ‘red pill right’1 who have attached themselves to the GamerGate movement’s tropes and memes.2 Indeed, many feminist and progressive stakeholders would resist the framing of the controversy and related events (including the violent harassment of female and progressive gamers, games researchers and game developers) around the term

‘GamerGate’, because that is the framing of the ‘pro-GG’ lobby. GamerGate is also a

‘born digital’ controversy with a strong presence across a number of areas of digital culture as well as different digital media platforms (from to message boards, from videogames to YouTube). GamerGate met our criteria for this study, then, both because it is sociocultural in its matters of concern, and because it is entangled with the politics and sociotechnical features of digital media platforms themselves (indeed,

1 http://boingboing.net/2015/01/28/a-beginners-guide-to-the-red.html 2 Teasing out these threads is beyond the scope of this article, but there are a range of primers 2 Teasing out these threads is beyond the scope of this article, but there are a range of primers with various ideological slants at sites such as Wikipedia and Know Your Meme, e.g. http://knowyourmeme.com/memes/events/gamergate Adrienne Massunari (2016) has convincingly explored the ways that Reddit’s specific architecture and culture works to support of the kind associated with

GamerGate).

From within the timeline so far of the overall GamerGate controversy (which is still ongoing), we selected a temporally bounded controversial event for the purpose of this analysis: an episode of the long-running NBC television drama Law and Order:

Special Victims Unit (hereafter referred to as SVU) episode that screened on 11

February 2015, and which was acknowledged by the show runners as being based on the gender-based harassment of female developers associated with

GamerGate. SVU promos frequently use the catchphrase ‘ripped from the headlines’ to indicate that the plots of the episodes are based on recent news events (Guittar, 2012).

Despite such claims to verisimilitude, the show is notorious for its unrealistic representations of young people, sexualities and technocultures, occupying the concerned (but clueless) adult perspective. The scripts are thick with expository dialogue designed to define and explain foreign concepts from online culture and other subcultures to mainstream audiences, and they often replicate generational stereotypes and media effects logics. Guittar (2012) has studied how SVU prompts misconceptions through a content analysis of episodes featuring LGBT hate crimes. He argues that the show’s framing of the issue can potentially lead to misperceptions about LGBT people, and stresses the lack of correspondence between current data on LGBT-based hate crimes and the offender profiles and motivations posited by the show (Guittar 2012:

220-21).

The ‘Intimidation Game’ episode is a more recent example of these trends. The episode was broadcast when GamerGate had already been unfolding for more than a year; during which a number of prominent female (or feminist) game designers, academics and critics had been harassed and threatened. In the episode, the detectives of the Special Victims Unit of the New York Department investigate a sexually related crime towards a successful indie game developer. This female character, Raina

Punjabi, who is being harassed online, is kidnapped from a conference and sexually assaulted by a group of young men devoted to videogames. Although the SVU detectives finally rescue Punjabi, she gives up and leaves the videogame industry saying that the harassers won the game.

The episode re-framed both the subculture and the central female protagonists/subjects of GamerGate in ways that were not accepted by the key players in the controversy, no matter which side(s) they were on. Gaming, technology and media news sites criticized the episode as being reductionist (Kain, 2015; Merlan, 2015;

Stark, 2015), defenders of the slogan GamerGate is about ‘ethics in game journalism’ argued that the episode was the result of a long press campaign of demonizing gamers

(Yiannopoulos, 2015), and feminist voices lamented the mainstream media lack of knowledge of the gaming industry, noting that the episode’s denouement represented female gamers as victims (Alexander, 2015; Merlan, 2015). Most interesting for the purpose of this paper, however, is the way that the episode’s outsider representation stimulated meta-reflection on GamerGate itself, bringing actors onto the stage and moving the controversy along.

Scope and limitations of the Twitter dataset

The background dataset for this study comprises tweets containing the keyword

‘gamergate’ (which includes but is not restricted to the #gamergate hashtag) posted from 17 November 2014 onwards. At the time of writing, this dataset contained more than 8 million tweets, with data collection continuing. A subset of tweets posted between 10-16 February 2015 (the period of acute activity including the screening and discussion of the Law & Order: SVU episode on 11 February 2015 at 9pm, New York time) was then extracted to begin the issue mapping process. The total number of tweets in the dataset was 238,967; and the total number of distinct Twitter accounts was 29,

278.

We acknowledge here that the choice to gather data under the hashtag

#gamergate in itself imposes a particular frame on the issues and sub-controversies involved – the #gamergate hashtag tends to be used far more by adherents to the ‘ethics in game journalism’ frame than it is by feminists and progressive gamers, game researchers and activists (for both political and safety reasons). Arguably, the creation and gaming of the hashtag by pro-GG actors is an example of a ‘forced meme’, inevitably skewing the visible public discussion towards the perspectives of the technoculture organized around ‘geek masculinity’ Massanari (2015) observes on

Reddit. Notably, a manual survey of the Twitter accounts belonging to key opponents and targets of the movement and suggests a deliberate avoidance of the hashtag even when directly referring to GamerGate activities.

At the time of data collection, we were restricted to data that had already been collected – only tweets containing ‘gamergate’ as a keyword are included in our dataset.

With more advanced capabilities for retrospective data collection, after identifying the

Law & Order episode as a key object of study as part of the issue mapping process, we could gathered data from the hashtags #gameonsvu, #lawandordersvu,

#intimidationgame and #svu for the week in question. With these hashtags, we may have captured a more diverse discussion about the episode (including from participants with little prior interest in #gamergate itself).

Further, we acknowledge the extensive organizing, debating and engagement with the controversy and related issues that were conducted in closed channels, but that our study has so far not being able to engage. This means that we have been able to capture only a portion of the publicly available data explicitly concerned with the controversy and that therefore we do not claim to be mapping the entirety of the intersecting issue publics engaged with it; rather there is an inevitable analytical focus on the activities and strategies of the proGG participants. This by no means implies a different valuation on our part of ‘public’ vs ‘private’ publics; rather it simply highlights the importance of mixed methods including the integration of participant observation and ethnography, however for the time being addressing this imbalance is beyond the scope of our approach.

There are additional minor technical limitations associated with the fact that

‘gamergate’ is at times an extremely high volume keyword, causing spikes that hit the rate limit on the TCAT server being used to collect the tweets. When this happens, all tweets above the API’s in-built rate limit are missed, but unfortunately there is no way of knowing exactly how many tweets are missed from any particular dataset.3

Building the Issue Inventory

In order to examine overall patterns of activity first we performed basic data processing to prepare the TCAT data for the visualization software Tableau, extracting URLs, hashtags, and @mentioned accounts, and sorting tweets into categories based on their addressivity (plain tweet, @reply/mention, or retweet), before focusing on inventorying the hashtags used (as a proxy for topics and themes). At the time the SVU episode was screened in the US there was a peak in activity on the #gamergate hashtag, and a

3 This has occurred on average around twice each month for the period in which GamerGate tweets are being collected, so there is an unknown number of spikes in the dataset with an artificially low ‘ceiling’, and we are unable to determine how much ‘height’ has been cut off. prevalence of SVU-related ‘ad hoc’ hashtags (Bruns & Burgess, 2015), such as

#gameonsvu, #lawandordersvu, #svu and #intimidationgame, indicate strong levels of engagement with the episode as it went live to air (see Figure 1). The following day we find another peak in activity with users still discussing the episode.

Two persistent hashtags broadly used by GamerGate proponents, #notyourshield and #opskynet, were visible throughout the week, with a peak in the use of

#notyourshield the 15 February. Likewise, among the hashtags most-used in combination with #gamergate, we identified in-jokes such as the use of the hashtag

#fullmacintosh to mock Jonathan McIntosh, the producer and co-writer of Anita

Sarkeesian’s Feminist Frequency, terms related to the game industry (#gamedev,

#video_game_tester_jobs, #games, #gaming) and unrelated terms

(#daylightsavingtime), some of the latter probably being the result of ‘hashtag jumping’

(Christensen, 2013), a common practice of using marketing bots or spam accounts to hijack trending topics on Twitter.

Figure 1: Most-tweeted hashtags over time (without #gamergate) Top 20 most-tweeted hashtags from 10-16 February out of 4110 hashtags.

As a first step in exploring the media objects that emerge as being significant across platforms (e.g. mainstream media news items, meme images, YouTube videos) and act as key mediators in the discussion around #gamergate, we filtered our Twitter data to identify the URLs that were most-shared throughout our week of analysis. The story that circulated the most during that week is an article4 by Brianna Wu, one of the targets of the GamerGate campaign, published on 12 February in Bustle entitled ‘I’m Brianna

Wu, And I’m Risking My Life Standing Up to Gamergate’. This link was originally tweeted by 406 distinct users and retweeted 1,913 times. The second most shared URL was an entry of the site blogjob.com entitled: ‘How Wikipedia uses false information to defame #gamergate’, followed by a URL to the dossier ‘A Review of Game Journalist’, a broken link at the time of writing although by means of a quick search on one can locate copies5 of that dossier that users uploaded before the original link went down.

Other relevant media objects were articles about the GamerGate controversy published in different websites, from Buzzfed to pro-GG websites, and one article published on

Forbes about the SVU episode ‘Intimidation Game’. 6

Second, we proceed to investigate the role of other platforms during the week the SVU episode was released. We filtered the URLs shared by domain (see Figure 2) and identified the top 15 relevant sources of information. Twitter.com prevailed over the others, followed by YouTube, the webpage capture tool Archive.is, the game related websites Bustle.com, Gamingjobonline.com (a spam link), Blogjob.com and Breitbart; and significantly, message boards Reddit, , and the image-sharing site Imgur, all

4 http://www.bustle.com/articles/63466-im-brianna-wu-and-im-risking-my-life-standing-up-to- gamergate 5 http://i.xomf.com/ypjxw.png 6 http://www.forbes.com/sites/erikkain/2015/02/12/law-order-svu-takes-on-gamergate-cant- press-reset-button/ three of which indicate the embeddedness of GamerGate within the irreverent, frequently male-dominated spaces of popular internet culture.

Figure 2: 15 most-tweeted domains after URL resolution Top 15 most-tweeted domains from 10-16 February out of 1429 domains.

This first data exploration helped us to identify important hashtags and other media objects, such as YouTube videos, in the discussion around #gamergate, and gave us a sense of the media ecology within which the Twitter activity is more broadly embedded.

The next step was to map the links of association among these media objects and the issue networks that emerged from these connections.

Mapping the issue networks

While in some hashtag-based Twitter studies it is useful to construct a social network graph based on either @replies or follower-followee relationships between user accounts, for the purposes of this research we are focusing on the networks of associations among media objects as key mediators in the issue; and using these to identify and understand the relationships between the diverse discursive frames in play.

We first extracted from our Twitter dataset a co-hashtag network file, which we visualized as an undirected graph where the edges (or connections) represent co- occurrence within tweets. That is, the more frequently any two hashtags appear together in a tweet, the stronger the link between them.7 A large cluster was identified, containing directly SVU-related hashtags such as #gameonsvu, #lawandordersvu,

#intimidationgame and #svu, which often co-occurred with the pro-GG hashtags

#notyourshield and #opskynet. From within these pro-GG hashtags we identified a specific anti- cluster containing the hashtags #womenagainstfeminism,

#feminismisawful, #antisjw and #feminism (See Figure 3). Another significant cluster revolved around the International Festival of Independent Gamers with #indicadeest co- occurring with hashtags such as #indiedev, #, # and #Journalism— representing a focus on games culture and the ‘ethics in games journalism’ framing that many pro-GG protagonists insist on.

Figure 3: Detail of anti-feminism cluster (concentrated on the right) (without #gamergate) Directed graph based on co-word analysis of hashtags with 244 nodes and 12625 edges. Size encodes word frequency. We applied the Giant component filter, which

7 We set a minimum frequency of 20, producing a list of hashtags that co-occurred at least 20 times during the week of analysis. keeps only nodes in the most important cluster. Before the filter: 3855 nodes and 1265 edges.

Because YouTube stands out as the second most used domain by the users of

#gamergate on Twitter, we decided to zoom in into this platform and map the relations between the videos that were shared on Twitter. In order to examine the discursive communities on YouTube based on the videos that were shared on Twitter, we used the

YTDT Video Network tool (Rieder, 2015a) that generates a network of relations between videos (based on the results of the YouTube ‘related videos’ algorithm). We filtered the URLs in Tableau for all links that contained the string ‘’, extracted the YouTube Ids and input them into the tool. We obtained a network8 file for the URLs that we extracted from a particular timeframe of the #gamergate hashtag on Twitter, which we visualized with Gephi. First, we generated a graph using the algorithm modularity class, which divides the dataset in different communities based on shared thematic affinities and interaction. For this first exploration we focused on the videos that were interconnected and ignored the videos that were shared on Twitter but emerged as satellites on the graph. As a result, Figure 5 shows the ‘controversy clusters’ of videos discussing GamerGate. As shown in Figure 5, the first big cluster in pink comprises videos explaining what GamerGate is, videos debating the SVU episode and videos arguing what critics of the movement get wrong about GamerGate. In green there is another community focused on the critics of GamerGate that advocate that the movement is about misogyny. In blue we found a cluster mostly dedicated to the cultural critic Anita Sarkeesian, another of the targets of GamerGate, with videos commenting on different topics, such as the decision of the Anti Defamation League to

8 We set the parameters to not crawl further than the ~350 seeds prepare teaching materials based on the work of Sarkeesian and aimed at discussing manifestations of sexism and misogyny in the gaming world, and videos discussing how the videogame Towerfall Ascension added a new a playable character based on her.

Figure 4: Gamergate communities on YouTube Directed graph with 354 nodes and 1102 edges. Size encodes view count and color defines the different communities after applying the algorithm Modularity.

Identifying the key mediators

By mapping the video networks we discovered themes that are salient to the GamerGate discussion (i.e. gender politics in videogames) and that extend beyond the SVU episode.

In order to identify relevant media objects in the #gamergate controversy for closer analysis, we focused again on the videos shared on Twitter and we aimed to identify which ones acted as key mediators on YouTube. For this purpose we generated a second visualization with Gephi using the same video network file but this time applying a betweenness centrality algorithm. This metric identifies the nodes that act as ‘bridges’ between ‘communities’.

In Figure 5 below, node size encodes the number of overall views and color the degree of betweenness--the nodes in red are those with the highest degree of betweenness. The cluster arranged in the middle is highly connected and constitutes the

‘controversy cluster’ associated with GamerGate. Where the nodes are media objects

(video texts), a high degree of betweenness within an overall context of ideological diversity indicates that they act as a site of intersection and potential co-awareness for the diverse and competing perspectives across the GamerGate video network, prompting video responses in support or rebuttal, and keeping the ‘controversy cluster’ dense. Our qualitative analysis of this cluster suggests that orthodox pro-GG perspectives have the highest betweenness. Examples include videos entitled

‘#gamergate in 60 seconds’ and ‘Angry Joe TotalBiscuit and #gamergate’. Also in this cluster are: one video associated with the ‘quinnspiracy’ frame (entitled

‘InternetAristocrat’s #gamergate series’); and two videos commenting directly on the

SVU episode. Another key mediator is a video that responds to one of the videos posted on Sarkeesian’s YouTube Channel Feminist Frequency, entitled ‘Tropes vs Women

Refund: An Open Letter to Feminist Frequency’; and a parody of feminists9. Most of these videos have more that 50.000 views on YouTube, and accumulate from hundreds to thousand of comments from diverse ideological perspectives.

On the contrary, as is also shown in Figure 5, the blue nodes around the

GamerGate cluster correspond to videos that are hugely popular in YouTube (some have hundreds of millions of views) but that are not interrelated between them or to the

9 The video is entitled ‘#gamergate sings: We Do (it for free)’ and had 525 views at the time of writing. GamerGate YouTube network. These are videos that users have used as annotations or humorous comments (i.e. MC Hammer videoclip ‘U can’t touch this’), acting as footnotes to the conversation. This network visualization helps us to identify the media objects that are interesting for a closer qualitative analysis, as well as for following across to other platforms.

Figure 5: YouTube video network Directed graph with 354 nodes and 1102 edges. Size encodes view count and color defines the degree of betweenness. The nodes in red are the ones that appear most often on shortest paths between nodes in the network.

To further understand the relevance of these media objects in the discussion of

GamerGate across platforms, we studied to what extent these YouTube videos also circulated on . We used the software tool Netvizz (Rieder, 2013) to search how many times the top videos originally posted by different users on Twitter had been shared, liked and commented on Facebook by the date on which the tool was run in

November 2015, producing a table with the share, like and comment counts on Facebook (see Table 1). The results indicate that these media objects also have relevance on Facebook and can be considered acute sites of controversy, since some of them had been shared more than 2,000 times and commented more than 1,000 times.

Table 1: Facebook circulation as at 13 November 2015 of the 10 most-shared YouTube videos on Twitter. Video Title YouTube URL Shares Likes Comments Totals GamerGate - If It's Not youtube.com/watch?v= 224 259 56 539 About Ethics... wy9bisUIP3w #gamergate in 60 youtube.com/watch?v= 2608 824 1382 4814 Seconds ipcWm4B3EU4 #gamergate - A youtube.com/watch?v= 506 414 347 1267 Reminder JCSZ0gUfbMg #gamergate The youtube.com/watch?v= 115 44 110 269 Movie RED BAND 7byUpdCUtuk Trailer (2016) - What is GamerGate? The Future Of youtube.com/watch?v= 28 21 30 79 #gamergate! -- Indie- z0g3rz75HPU Fensible Another Dramatic youtube.com/watch?v= 42 98 20 160 Reading of Anti- 0N_xoa8r2Nk GamerGate Tweets | #gamergate Why We Must Take youtube.com/watch?v= 369 822 852 2043 Action- The Porn 9Z85GQF9--s Charity, Mercedes Carrera Gamergate and the youtube.com/watch?v= 0 0 0 0 War on Fun aBZrmW3tMGs Another Game youtube.com/watch?v= 163 106 84 353 Developer Answers EN7Qy9N1C9E More of Your Questions! Law & Order SVU youtube.com/watch?v= 0 0 0 0 Takes on Gamergate n7faUHdlh9g [removed due to copyright claim by NBC Universal]

Overall, while most of the videos that were posted by the highest number of distinct users on Twitter also act as key mediators on YouTube (videos with high degree of betweenness) and Facebook (high number of shares and comments), some of these media objects operate differently. For instance, the video entitled ‘GamerGate and the war on fun’ was posted by two distinct users and retweeted 59 times on Twitter, whereas on YouTube it only has 24 comments and 924 views, and it was never posted on Facebook. This is a first indication that although media objects are shared across platforms when controversies are discussed, some key mediators in one platform will not necessarily be relevant in other platforms, which unfolds the importance of accounting for the cultural specificity of different digital media platforms.

Nonetheless, some media objects are definitely key mediators in the dynamics of communication across platforms. The obvious example is the #gamergate hashtag itself, which not only plays a role on Twitter, but orchestrates the discussion around this controversy in other platforms, such as Instagram, Facebook, YouTube and Tumblr. To test this hypothesis we followed the ‘gamergate’ tag on Tumblr, and examined the tag networks that emerged by querying #gamergate using the DMI Tumblr tools, again focusing on the week that the SVU episode screened (Rieder, 2015b). The visualization in Figure 6 indicates that the GamerGate controversy is not only a two-sided debate about a shared issue, but a site of intersecting publics with diverse interests and political perspectives. In red we identified tags associated with positions about GamerGate that disagree with each other (i.e: Law & Order SVU, Zoe Quin, Feminism, Gamers,

Equality, etc ), in blue we have tags commonly used in digital culture to discuss the politics of feminism and gender (i.e. misandry, maleprivilege, , yesallwomen), and in purple there are tags revolving around other configurations of contemporary identity politics and social justice issues (i.e. LBGT, sexism, human rights, reverse , racism, etc.). On Tumblr, the #gamergate hashtag affords the collision and mutual awareness (however marked by antipathy) of the diverse constituencies of intersecting publics—from the games industry to gamer culture to television fandom to gender and identity politics.

Figure 6: Tumblr tag issue networks Undirected graph with 994 nodes and 6789 edges. Size encodes degree and color modularity.

Reflections on Findings

Because we gathered data from #gamergate (which is mainly used by pro-GG participants) the issue networks identified in this study correspond mostly to pro-GG views. Our findings show that, even when initially approached from as partial a perspective as the ‘gamergate’ keyword and hashtag represents, GamerGate’s issue publics are absolutely not concerned only or even primarily with ‘ethics in games journalism’; despite the fact that debates around the Law & Order: SVU episode were very concerned to argue that the show misrepresented GG and that this (ethics in games journalism) was what the movement was actually about. In a finding that will be no surprise to feminist games scholars and gamers, we identified actors within the

GamerGate movement that use the hashtag only to denigrate or harrass women—both key figures such as Sarkessian, Wu, and so on, and women in general. We have also identified key media objects (from hashtags, to videos and tumblr posts) encoding anti- feminist perspectives. Pro-GG media objects are used to mock women and/or critique feminism across a range of digital media platforms, from YouTube videos to Tumblr and Reddit posts. By undertaking a further cross-platform analysis and a future qualitative analysis it would be possible to more richly describe the expressive and discursive practices through which such misogyny unfolds (through humor, harassment, etc); and the way they articulate to the specific technocultures of particular platforms

(Massanari, 2015). Relatedly, our results point to the expertise of GG as a technoculture in shaping issues and algorithmic curation (e.g., the evidence of their use of bots on

Twitter to promote the hashtag #gamergate, their use of the imgur image-hosting service to post text in a .jpg format, the use of the capture tool Archive.is).

Our results indicate that through closer qualitative analysis it would be possible to highlight alternative and opposing perspectives even within the issue publics associated with the ‘gamergate’ tag. For instance, among the Tumblr posts we analysed was a critique of key GG protagonist Adam Baldwin’s , potentially engaging all Tumblr users participating in the gamergate discussion by using the

‘gamergate’ tag on the post. Future developments in issue mapping via social media may allow us as researchers to highlight and make visible the expressions of these counter-publics, thus amplifying diverse and minority perspectives in order to enrich the public debate (while being careful not to compromise their authors’ and curators’ safety).

Conclusions and forward agenda

Through a study of a sub-controversy within the ongoing GamerGate phenomenon, this paper demonstrated an empirical approach to accounting for the role of digital media platforms and cultures in shaping public issues, and the issue publics that engage critically with those platforms. The paper demonstrated the importance and also the considerable challenge of truly engaging with the sociocultural dynamics (from race, gender and sexuality to humour and celebrity) that are ever-present in everyday social media communication, and therefore are important in the life of issues and their publics.

The key contribution of the paper is our novel focus on the roles of key media objects

(images, videos, and tags) in coordinating and progressing the controversy. The

#gamergate example highlights the importance and challenge for research-at-a-distance of engaging with the cultural genealogies and politics of the issues and controversies under study.

The greatest benefit of the issue mapping approach for studying digitally native controversies, however, lies in its ability to get beyond the loudest voices and binary oppositions, to reveal the multi-sidedness and intersectionality of social media controversies. Advanced combinations of network and content analysis can be used to identify minority perspectives and to highlight those media objects with high degrees of betweenness (that act as bridges between distinct network clusters); and especially to identify the clustering patterns and interactions among diverse interest-based communities and issue publics (e.g. among gender politics and game cultures). While we have touched on such questions here and some early work has been done on solving the problem (Bruns & Burgess, 2015; Marres, 2015; Marres & Moats, 2015) issue mapping methods are still only at the beginning of critically acknowledging and empirically accounting for the sociotechnical agency of platform architectures and activities such as the algorithmic curation of trending topics, hashtags, and hence the algorithmic shaping of issue publics. We note the way this case study has pointed to the sociotechnical shaping of issue publics, not only ‘passively’ (e.g. via the algorithmic ordering of search results), but also actively, as when pro-GG participants are aware of and bent on exploiting the affordances and governance gaps in digital media platforms to advance their agenda.

There remain at least two further key methodological challenges that this paper only opens up but does not yet solve:

First, the challenges of analysing multimedia content across platforms. We are only at the very early stages of developing truly integrated, multi-platform approach to data collection and analysis, expanding well beyond Twitter research—which has by now developed a standard set of orthodox approaches to gathering and analysing data from the Twitter API—to incorporate image-sharing, video and rich content curation platforms like Instagram, YouTube and Tumblr. There is much more to do on integrating into digital media research techniques like cross-platform media object tracking (Driscoll & Thorsen, 2015) as well as large-scale classification and close visual content analysis (Thelwall et al, 2015). as well as to operationalize issue mapping taking other social media platforms as the starting data sources, such as Reddit discussions around a given topic (scraping comments, URLs, images and reproducing similar analysis). Second, the challenges of integrating public engagement throughout the project and at every level of the methodology. Consideration needs to be given to ways of publicizing interim findings and results that allow audiences to quickly and easily gain an overview of the key issues of a debate and the stakeholders involved without over- simplification. Such activities will enable researchers to observe the way publics themselves use digital methods and data science as part of their engagement in issues— as has been the case with GamerGate, where basic metrics and sentiment analysis of

Twitter data was deployed by corporate analysts in service of attempts to determine whether the movement was ‘really’ more about harassment than ethics in games journalism, finding that it looked very much like harrassment (Wofford, 2014); while others used informal content analysis to ‘prove’ that not only was GamerGate indeed about ethics in games journalism, but that male participants in the hashtag were the ones being harassed. Academic researchers with digital methods at their disposal have a significant opportunity and responsibility to intervene in such debates, and it is vital that we continue to develop methods that are up to the task.

References

Alexander, L. (2015). Law & Order: SVU ‘Intimidation Game’ is not what games are about. Hopes and Fears. http://www.hopesandfears.com/hopes/culture/video- games/168413-gamergate-svu-episode-reviewed-by-leigh-alexander Berlant, L. (1997). Introduction: the intimate public sphere. The Queen of America Goes to Washington City, Essays on Sex and Citizenship, 1-24. Bruns, A., & Burgess, J. (2011). Gawk scripts for Twitter processing. Mapping Online Publics. http://mappingonlinepublics.net Bruns, A., & Burgess, J. (2015). Twitter hashtags from ad hoc to calculated publics. In Rambukkana, N. (Ed.) Hashtag Publics. New York: Peter Lang. (in press) Burgess, J., & Crawford, K. (2011). Acute events in social media. Paper presentation, IR12.0: Association of Internet Researchers Conference, Seattle, Washington, USA. Burgess, J., Galloway, A. & Sauter, T. (2015). Hashtag as hybrid forum: the case of #agchatoz. In N. Rambukkana (Ed.), Hashtag Publics. New York: Peter Lang (in press, available at http://mappingonlinepublics.net). Callon, M., Lascoumes, P., & Barth, Y. (2001). Acting in an Uncertain World: An Essay on Technical Democracy. Cambridge, MA.: MIT Press. Christensen, C. (2013). Wave-riding and hashtag-jumping: Twitter, minority ‘third parties’ and the 2012 US elections. Information, Communication & Society, 16(5), 646–666. Dahlgren, P. (2009). Media and Political Engagement: Citizens, Communication, and Democracy. Cambridge: Cambridge UP. Driscoll, K., & Thorson, K. (2015). Searching and clustering methodologies connecting political communication content across platforms. The ANNALS of the American Academy of Political and Social Science, 659(1), 134-48. Fraser, N. (1990). Rethinking the public sphere: A contribution to the critique of actually existing democracy. Social text, (25/26), 56-80. Gibbs, M., Meese, J., Arnold, M., Nansen, B., & Carter, M. (2015). # Funeral and Instagram: death, social media, and platform vernacular. Information, Communication & Society, 18(3), 255-268. Gillespie, T. (2010). The politics of ‘platforms’. New Media & Society, 12(3), 347–364. http://doi.org/10.1177/1461444809342738 ______. (2014). The relevance of algorithms. In T. Gillespie, P. Boczkowski, & K. Foot (Eds.), Media Technologies: Essays on Communication, Materiality, and Society (pp. 167-94). Cambridge, MA.: MIT Press. Guittar. N. (2012). ‘Ripped from the headlines: The depiction of sexual orientation- based hate crimes in television crime dramas.’ In D.L. Bissell & J.L. Conners (Eds.) The Harms of Crime Media. (pp. 209-223). Jefferson, NC: McFarland. Kain, E. (2015). ‘Law & Order: SVU’ Takes On #gamergate, Can’t Press Reset Button. 11 February. http://www.forbes.com/sites/erikkain/2015/02/12/law-order-svu- takes-on-gamergate-cant-press-reset-button/ Livingstone, S. (2005). On the relation between audiences and publics. In S. M. Livingstone (Ed.), Audiences and Publics: When Cultural Engagement Matters for the Public Sphere (pp. 17–41). Bristol: Intellect Books. Marres, N. (2015). Why map issues? On controversy analysis as a digital method. Science, Technology and Human Values. [online before print doi: 10.1177/0162243915574602] Marres, N. & Moats, D. (2015). Mapping controversies with social media: The case for symmetry. Social Media + Society 1(2): 2056305115604176 Massanari, A. (2015). #gamergate and The Fappening: How Reddit’s algorithm, governance, and culture support toxic technocultures. New Media & Society. Online first doi: 10.1177/1461444815608807 Merlan, A. (2015). Gamergate and its victims hated the Law & Order Gamergate episode. Jezebel. 2 February. http://jezebel.com/gamergate-and-its-victims- hated-the-law-order-gamerga-1685386251 Papacharissi, Z. (2010). A Private Sphere: Democracy in a Digital Age. Cambridge: Polity Press. ______. (2015). Affective Publics: Sentiment, Technology, and Politics. Oxford: Oxford University Press. Rieder, B. (2013). Studying Facebook via data extraction: the Netvizz application. In Proceedings of the 5th Annual ACM Web Science Conference (pp. 346–355). ACM. Retrieved from http://dl.acm.org/citation.cfm?id=2464475 Rieder, B. (2015a). YouTube Data Tools. Computer software. Vers. 1.0, 5 May 2015 https://tools.digitalmethods.net/netvizz/youtube/ Rieder, B. (2015b). TumblrTool. Computer software. 3 July 2015. https://tools.digitalmethods.net/netvizz/tumblr Rogers, R., Sánchez-Querubín, N., & Kil, A. (2015). Issue mapping for an ageing Europe. Amsterdam: Amsterdam University Press. Rogers, R., & Marres, N. (2000). Landscaping climate change: A mapping technique for understanding science and technology debates on the World Wide Web. Public Understanding of Science, 9(2), 141–63. Stark, C. (2015, February 12). ‘Law and Order: SVU’ takes GamerGate to extreme to show ‘this could go too far’. http://player.theplatform.com/p/NnzsPC/widget/select/media/IBhOTKRrX0HE Thelwall, M., Gorunova, O., Vis, F., Faulkner, S., Burns, A., Aulich, J., Mas-Bleda, A., Stuart, E., & D’ORazio, F. (2015). Chatting through pictures? A classification of images tweeted in one week in the UK and USA. Journal of the Association for Information Science and Technology [pre-print: http://www.scit.wlv.ac.uk/~cm1993/papers/ChattingThroughPictures_preprint.p df] Todd, C. (2015). COMMENTARY: GamerGate and resistance to the diversification of gaming culture. Women's Studies Journal, 29(1), 64. Venturini, T. (2010). Diving in magma: How to explore controversies with actor- network theory. Public Understanding of Science, 19(3), 258–73. ______. (2012) Building on faults: How to represent controversies with digital methods. Public Understanding of Science, 21(7), 796-812. Warner, M. (2002). Publics and counterpublics. Public culture, 14(1), 49-90. Wofford, T. (2014). Is GamerGate about media ethics or harassing women? Harassment, the data shows. . 25 October. http://www.newsweek.com/gamergate-about-media-ethics-or-harassing-women- harassment-data-show-279736 Yiannopoulos, M. (2015, February 18). Now The Gaming Press Is Turning On Developers. Retrieved 23 November 2015, from http://www.breitbart.com/london/2015/02/18/now-theyre-coming-for-you- games-press-turns-on-developers/