When the World Ends: Study of #FakeWesteros

Introduction

On May 19, 2019, the 8-season run of HBO’s Game of Thrones (GoT) came to a much-anticipated and much-maligned end. A lot has been written by the entertainment press, critics and fans about what Benioff and Weiss, creators of the television show, could-have/should-have/would-have done and how they failed to bring this decade-long journey to a satisfying conclusion for all.  Much of the mainstream reaction to the final season was negative– a reflection, perhaps, of the profound engagement and psychological investment felt by many of the over 44 million viewers who watched it.

There are fans, and then there are fans. I know in this case that I am the former: long-time reader of ASOIAF, the book series that GoT adapted, and avid re-watcher of my favorite GoT episodes, my engagement with the show does not extend to community and participatory activities like live-tweeting, watch-parties or cosplay at cons (not for lack of consideration, but a good faux fur is expensive). In fact, as a researcher I was impatient for the ending precisely because I wanted to see first-hand how fans (not just fans, but fans) reacted.

I wanted to know what happens to fans when the story ends.

#FakeWesteros is a community of GoT fans on Twitter. These are fans whose engagement moves beyond the television screen: they are transmedia fans. Besides meeting this principal criteria for my study, the FakeWesterosi (or Twitterosi) bring with them a history and a set of unique performative, social and textual poaching practices that are shaped by the affordances of Twitter as a medium. “Livetweeting”episodes (i.e., commenting during viewing) from around the world, they are engaged in a half-joking/half-serious communal roleplay, assuming parodic versions of characters from GoT. The community has been around nearly as long as the show; it formed around the show, and now, with the show ending, the question of whether or not it would dissolve once more seemed to preoccupy the members, especially as the finale approached and mainstream ire grew. The ending of the tv series posed an existential crisis for #FakeWesteros, offering its own kind of drama: will they or won’t they…?


RELATED DOCS

CONTENTS


Questions

  • What is the reaction to the ending of the series?
  • What are the unique practices and behaviors of the #FakeWesteros community (fan practices, information practices) ?
  • How does the community negotiate information from each episode, from fan works and extension, from each other and the greater GoT fandom, from critics, and from the mainstream response?

Previous Work

This study builds on previous research regarding the GoT fandom and the information behavior of transmedia fans. In an earlier pilot study (Forcier, 2017), which explored the GoT fandom on the website The AV Club, I developed a preliminary model for the information behavior of transmedia fans, which established a cycle for the negotiation of narrative information. Analysis of #FakeWesteros can help test and improve this model in the development of a theory for transmedia fandom. The study of the #FakeWesteros community also fits into the larger tapestry of my PhD research, which aims to define exactly what it means to be a transmedia fan.

Rebecca Williams’ robust work mapping the moment when fans encounter the “beginning of the end” of their fandom, or at least, their fan object, provides a profound exploration of the phenomenon she refers to as “post-object fandom” (2015). Her book on the subject has effectively answered the question of “what happens to fans when the story ends” with examples from fandoms spanning a period from 2004 to 2013. In studying #FakeWesteros, I aim to further contribute to this body of work with results from a current instance and interpreted from the perspective of information behaviour.

Theoretical Perspective

This research is informed by a social constructionist paradigm and employs a constructivist grounded theory methodology (Charmaz, 2014), which permits an inductive and recursive development of ideas and theoretical models through the analysis of data. A preoccupation with the specific activities of individuals and communities that demonstrate how information is encountered and shared (i.e., information behavior) grounds this study firmly in the empirical.

Methodology

The study identified members with the hashtag #FakeWesteros in their profiles or that appeared in user-created lists, then generated a list of the 50 most active Twitter users during the airing of season eight from 12 April to 20 May 2019. Using the Twitter API, I collected the timelines for members during the above period, representing a total of 27,775 tweets.

Data collection and analysis considered air dates and times. The earliest airing was on HBO Sunday nights at 21:00 Eastern Standard Time (EST). While the episode was available in simulcast in other time zones and on other networks (for example, the simulcast on Foxtel aired episodes at 11:00 AEST on Mondays) and was also available on NOW TV and HBO Go streaming services after that time, the episode officially aired again on Sky Atlantic Monday nights at 21:00 GMT. Livetweeting, that is, member’s posting their reactions as they watch the episode, generally took place at one or both times– indeed, the most active users would livetweet multiple viewings of the same episode. Livetweet sessions were often immediately followed by commentary within the #FakeWesteros community. For this reason, data was collected and sampled to ensure full livetweet sessions were captured.

The study employed a dual approach for analysis of Twitter data:

  1. Qualitative coding, as described by Charmaz (2016) and used previously for the analysis of The AV Club comments (Forcier, 2017), identified key themes. As a sampling strategy, analysis compared tweets posted on the premiere and finale episode air dates on HBO and Sky networks (April 14 and 15 and May 19 and 20). The sample coded 10 out of every 100 tweets, for a total of 510 tweets.
  2. Natural language processing (NLP) using the Python VADER Sentiment Analyzer by Hutto and Gilbert (2014). Tweets posted within 7 days of the HBO air date were evaluated to determine a sentiment score that indicates the overall reaction to each episode.

Results and Discussion

A key theme identified tweets that conveyed a keen awareness of the impending end to the series in relation to a concern for its impact on the community (Fig. 1). This concern was expressed universally among members in their tweets in a variety of ways: earnest expressions of mourning for fallen characters (Fig. 2) and fear that real friendships might fade (Fig. 3) appear alongside posts that invited fellow members to continue posting and sharing new interests after the end of the series (Fig. 4). The use of humor and pastiche, involving the intertextual layering of information found in previous research (Forcier, 2017), served to “soften” what, for many members, was a deeply emotional experience.

Figure 1. @iMissandei_ expresses her hope that social bonds formed online will outlast the end of the series.


Figure 2. @TheLadySansa responds to @IronBornTheon in reference to the redemptive death of Theon Greyjoy earlier in the season.

 

Figure 3. @AerysGoneMad expresses heartbreaking emotion over the impending end, prior to the finale.


Figure 4. @ellariasnake starts a thread to plan book club, a new community activity.

Another theme that emerged was the use of graphic-based content as an essential form of expression. Figures 1-4 (above) provide examples of this. Emojis, memes, animated GIFs and intertextual references (Fig. 5) are all part of the fandom’s toolkit, and Twitter as a platform facilitates this multimedia communication. #FakeWesteros tweets are multi-layered: just as the roleplaying conceit layers character over real-life identity, visual media layers over text, changing meaning and interpretation. The layering of multimedia content here is far more prevalent than found in among The AV Club comments in the pilot study of GoT fandom, which suggests a transformation in how online communities share information. Further research would be required to determine exactly what factors have contributed to this change.  The use of multimedia also poses a challenge for NLP methods and highlights the need for further research to develop sentiment analysis approaches.


Figure 5. @LordGendry comments on a moment in “Winterfell” (episode 1 of season 8) with a screenshot and a reference to Harry Potter.

The “livetweet” session, as a type of fan-specific information practice, represents the unique production offered by the community. A set of specific behaviors, or tactics,  make up this practice: “quoting”, “reviewing”, “predicting” and “debating” are all behaviors observed that represent different aspects of each member’s livetweet of an episode. During viewing, favorite lines are quoted and scenes are referenced, sometimes with comedic or emotional effect. After the episode, members review in earnest, first revisiting favorite moments, then debating the significance of events and what it might suggest about future episodes. This practice is essentially performative. Members react to the show and offer commentary to an audience of followers on Twitter.  While community members interact with each other, often in-character, they also have interactions with other GoT fans who are not strictly associated with or are merely followers of #FakeWesteros.

In the pilot study of fan comments on The AV Club (Forcier, 2017)fan reactions were categorized into four types of tactics: reasoned, relational, comic and sentimental. This categorization was further tested in the analysis #FakeWesteros. In livetweet posts, the analysis identified a new category for the negotiation of narrative information that is challenging to classify with the act of “quoting”. Tweets that directly referenced specific lines or scenes from the show shared characteristics of the four tactics for negotiating narrative information, but also marked differences. In The AV Club posts, such references included qualifying commentary that indicated how the user was negotiating information from the episode, with expressions of strong emotion (i.e., sentimental negotiation), rationalization or justification (i.e., reasoned negotiation), and/or comparisons with other stories, fandoms, texts or paratexts (i.e., relational negotiation).  The majority of #FakeWesteros tweets demonstrated one or more of these tactics, even when commentary included no text (Fig. 6). Some tweets that represented “quoting”, however,  did not include any  qualifying commentary (Fig. 7). There are several potential reasons for this. For one, comments analyzed in the pilot study were posted after The AV Club members viewed the episode, while #FakeWesteros livetweets represent reactions during viewing. Tweets also include less text, overall. The tweet itself, as a performance, mirrors the sentiment in the scene that it apparently mimics, but may also be highly contextual, relying on knowledge of the community member’s past history to properly interpret. Further analysis with a larger sample of tweets that looks specifically at “quoting” as an information behavior, or tactic, would help determine how it informs the information behavior cycle model proposed in the pilot study.


Figure 6. @LordGendry demonstrates sentimental negotiation with a single heart emoji and an animated GIF of the reunion between Arya and Gendry.


Figure 7. @LordBranRaven  quotes Bran Stark from the finale.

Natural language processing methods are commonly used for sentiment analysis of Twitter data (e.g., Hutto and Gilbert, 2014). This approach is  appropriate to obtain an overall sense of the community’s reaction to the series end. This part of the analysis included the full dataset, including tweets posted on April 12 and 13 prior to the airing of episode one (Fig. 8). Only tweets flagged as English were included. The VADER (Valence Aware Dictionary for sEntiment Reasoning) lexicon was selected for this project, as human-curated gold standard sentiment lexicon, and as it is specially designed for microblog-like social media content. It was determined to be more accurate than similar lexicons (such as LIWC and SentiWordNet) (Hutto & Gilbert, 2014). VADER  incorporates sentiment-laden lexical features like initialisms (e.g., WTF, LOL) and common slang (e.g., meh, nah). Another benefit of the VADER lexicon is its emphasis on sentiment intensity,  which measures not just the binary polarity of a word (positive versus negative), but also the strength of the sentiment (e.g., exceptional vs. okay).  VADER assigns a numeric value to determine sentiment polarity on a continuous scale ranging from –1 to 1 (most negative to most positive). The study identified a range for the Compound Sentiment Score of tweets within 7 days of the air date (e.g., episode 1 = 14-21 April).

Figure 8. Compound Sentiment Scores per episode as a box plot graph.

Figure 8 indicates that the community’s reaction to episodes in tweets were mostly positive. For all episodes and including tweets on 12 and 13 April prior to the episode, the median score was 0.00 or greater, indicating that positive sentiment dominated in tweets. The first quartile (i.e., 75% of tweets) for the finale, Episode 6, scored above 0.00, indicating that fewer than 25% of tweets were negative. For tweets prior to Episode 1, Episode 1 and Episode 6, the most negative compound sentiment scores were above -1.00, indicating fewer intense negative reactions. When we take into consideration the mainstream negative reaction popularly reported around the season and especially the finale, these results are perhaps surprising. However, qualitative analysis also found that #FakeWesteros members responded positively, even protectively, aligning themselves in solidarity with the producers and performers in GoT. Additionally, the positive expressions and affirmations around the community itself, as seen in some of the examples above, would have affected Compound Sentiment Scores.

For a description of our VADER analysis, the repository for this project can be found at https://github.com/jaderjj/Vader-Sentiment-Analysis

Conclusion

Days before the series finale, user @IronbornTheon reflected on the end of GoT, writing: “while I’m both excited, dreading it, and sort of not wanting to see it, I have realized that just because the show is going off the air, doesn’t mean the community is going anywhere.”

The idiosyncratic activity of #FakeWesteros members began with “para-active engagement” through their performance online, and the negotiation of narrative information (Forcier, 2017). However, the content that they have produced online and the interactions they continue to have move beyond the model developed in previous research. The data collected for this case study represents the negotiation of social identities and interpersonal relationships in the digital space. @IronbornTheon’s post is an example of the engagement and commitment of these fans to finding a future for the community, independent of the narrative that brought them together.

As a case study, #FakeWesteros demonstrates how information is transmitted and transformed innovatively and multimodally in online fan communities. It further demonstrates how identity and loyalty shape collective practices and individual behaviours, exhibiting the transition from “fan/object pure relationships” to “fan-fan pure relationships” (Williams, p. 21).

This research is, of course, a work-in-progress. As I’ve pointed out throughout this narrative, there are many directions in which this study can be expanded. Additional analysis is required for theory development and to fully capture the information behaviours and negotiation tactics of community members.  Further work with the VADER lexicon and NLP methods could produce more meaningful results.

Acknowledgments

This research is supported by the Social Sciences and Humanities Research Council of Canada and by Swinburne University of Technology.  Special thanks to Riley Jefferson for assistance with Python and VADER.

References

Charmaz, K. (2014). Constructing Grounded Theory: A Practical Guide Through Qualitative Analysis, Second edition. Thousand Oaks: Sage.

Forcier, E. (2017). Re(a)d wedding: A case study exploring everyday information behaviors of the transmedia fan. Proceedings of the 80th Annual Meeting for the Association of Information Science and Technology. DOI 10.1002/pra2.2017.14505401011

Hutto, C.J. & Gilbert, E.E. (2014). VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Text. Eighth International Conference on Weblogs and Social Media (ICWSM-14). Ann Arbor, MI, June 2014.

Tufekci, Z. (2019). The real reason fans hate the last season of Game of Thrones. Scientific American. Retrieved May 17, 2019 from https://blogs.scientificamerican.com/observations/the-real-reason-fans-hate-the-last-season-of-game-of-thrones/

Williams, R. (2015). Post-Object Fandom: Television, Identity and Self-narrative. New York: Bloomsbury.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.