#GamerGate versus Other Popular Hashtags: Part 1, Grounding and Verification
NOTE: This post has A LOT of early context and process explanations. You can skim down to the first graph if you just want to read the analysis. Future posts will jump into the analysis now that the methods and rationale are up front.
A great deal of interesting data analysis about #GamerGate has hit the web over the last few days. You should check out what Andy Baio and Brian Keegan have done if you haven't already. Since I’ve been running some samples for future academic work, I thought it worth taking a moment to show a smidge of what I’ve seen in order to look at a few different angles and to offer a minor verification study of the other analyses. After all, I’ve been talking about these trends for awhile.
This post will look at the following:
- How does #GamerGate’s interactivity compare to other hashtags?
- How do the posting sources of #GamerGate compare to other hahstags?
- Who were the top mentions during the period of this study?
24 Hours of #GamerGate and Five Other Hashtags
In starting this analysis I took a 24-hour window over October 26th. This generated a smaller sample of #GamerGate than Andy Baio. My sample consisted of 50,828 #GamerGate tweets compared to Baio’s 316,669.
Why the smaller sample? Because I wanted to show #GamerGate in comparison to other hashtags and picked a manageable 24-hour range over Sunday October 26th (Baio looked at the more active period of October 21–23). Choosing a Sunday offered prime hashtag time with #NFL, #TheWalkingDead, and Sunday morning news shows about #Ebola and #ISIS.
Limitations
Let’s get a quick list of basic limitations out of the way.
- I only collected data from Twitter for a single day. This selection leaves other conversations spaces as yet unexplored.
- I only collected data using the hashtags listed. And there are a lot of @replies without the hashtag.
- I did not collect all tweets for Sunday due to technical limitations, but I did sample tweets throughout the entire day. (My collection process is explained below.)
A Quick Note on Bias
My intent is to strive for a rigorous and valid representation and that’s why the variety of comparisons come in handy. That said, I’ve been critical of #GamerGate for its origins and its governance style—both of which I believe contribute to the level of harassment that occurs around the hashtag. I fully admit my point of view and own any concerns my past and current criticisms might raise with this analysis.
The data is objective, but my questions are rhetorically constructed. Frankly, this is true of anyone and to deny it is misleading in my opinion.
Secondly, this analysis is much less detailed than what I plan to publish. Posting it here is a personal mission to share information and analysis with the public outside of the stylistic and access barriers of an academic journal. Due to this, what I post here is not sponsored research in any manner. This analysis represents work I have done on my own time and reflects neither my affiliation with Texas Tech University nor my affiliation with MIT.
Also, I don’t know anyone personally who has held a high profile position in #GamerGate, but I do have a few minor professional relationships in this ever-expanding issue. I did an interview with Mike Mearls and Jeremy Crawford for The Mary Sue, which is a site sometimes cited in the Gamers are Dead articles list. I’ve also interviewed Felicia Day for SXSW and we attended the same university at the same time (I never met her while at Texas. Hook ‘Em.).
To mitigate concerns of bias, I’m going to remain largely descriptive in this first post. However, other posts in this series will attempt to draw some more specific conclusions as the series continues.
On with the show.
The Basic Process
I used DMI-TCAT to gather the tweets and perform basic analysis from Twitter’s streaming API. I then performed additional second level analysis via Excel, SPSS, and Gephi (for later posts). The data was collected from the 26th through the morning of the 27th. Since DMI-TCAT relies on chron commands to run data gathering, I collected a sample of tweets throughout the day and not all tweets appear in my data collection for that reason. I ran a grab every 15 minutes.
I compared the following hashtags: #GamerGate (50,828), #NFL (28,947), #TCOT (47,920), #TheWalkingDead (115,663), #Ebola (45,046), and #ISIS(28,171).
Picking #NFL over the non-hashtag “football” was a limiting choice. Not as bad as picking the abysmally used “#football” over “football” but still far from perfect.
And, yes, I literally compared #Ebola and #ISIS to #GamerGate. I offer you three ways to view this:
A) I’m engaging in hyperbolic rhetoric just to offend people.
B) I’m poking fun at said hyperbolic rhetoric.
C) These are three long-standing event-based hashtags in the public discourse and I’m not saying anything about the actual users but rather looking for trends in event hashtags.
A Peek at User Interactivity
Since I analyzed a variety of tweet totals among the hashtags, the following chart uses percentages to make comparisons accurate.
The chart above shows that RTs and mentions in #GamerGate, while higher than in other hashtags, aren’t too out of line with community hashtags like #TCOT. Where #GamerGate truly differentiated itself Sunday was in @replies, more than doubling second place #Ebola (5.85%) with 12.51% of #GamerGate tweets coming from @replies.
It appears that #GamerGate defies Twitter conventions with a much heavier dose of the hashtag in @replies than other conversations. It also appears to be more interactive across the board than the other hashtags examined on this day. These stats could speak to both the hashtag’s evangelical and community orientations.
Verification Side Note: My findings largely agree with Baio in RT percentages (69%) and @reply percentages(12.51%) even though we looked at different days.
The Workspace of #GamerGate
As a researcher in technical and professional communication, I’m quite curious about the workspace of #GamerGate posters. Brianna Wu posted two charts a couple of weeks back about the platform sources of #GamerGate posters. Wu believed that the heavy use of Twitter’s web client by #GamerGate versus #xbox was a sign of bot use. Many challenged the comparison she made, so I wanted to run a second test on #GamerGate posting platforms versus a variety of hashtags.
Twitter for Web Client is 11–25% for all other hashtags, but it jumps to 59% for #GamerGate in this sample.
What these charts all show is that #GamerGate absolutely uses Twitter’s web clients far, far outside the norm, as Wu stated. This goes well beyond alignment with any one hashtag and is a strong deviation from many different hashtags.
There are three likely reasons in my opinion.
- Demographic: For whatever reason #GamerGate users simply have less smart phone access and more PC access.
- Automation: Perhaps not exclusively bots, but some of the RTs in #GamerGate could be automated and driving the web client count higher than normal.
- #GamerGate is Semi-Professionalized: It is possible given the on-boarding documentation and determined nature around #GamerGate that those involved see it as work. If they are emailing advertisers, posting with evangelical intentions to Twitter (those @replies with hashtags), and contributing to documentation sources then it makes sense to do this from a work environment (like a laptop or desktop machine). Note that work could also mean that a core of professionalized posters dominate the total tweet counts and sway the demographics.
None of the above can be verified in this analysis, but 2 and 3 are compelling points for further exploration. In either case, Wu was correct in stating that there is an odd pattern in how #GamerGate posts to Twitter. However, the reasons for this odd behavior remain obscured.
Sunday the 26th #GamerGate Top 25 Mentions
For the last bit today, I’m going to look briefly at mentions in only #GamerGate because comparison does little good here.
Much like what Newsweek discovered, the #GamerGate hashtag on Sunday engaged little with journalists beyond Gawker and quite a bit with Sarkeesian and Day. It’s certainly likely that some of these are opponents of #GamerGate engaging with fellow critics, but that doesn’t explain the lack of criticized journalists appearing in the top 25.
An editorial by @suellentrop did gather attention on Sunday, but he’s the only journalist critical of #GamerGate in the top 25 and not a target for claims of ethical violations. It seems to suggest a continued lack of focus within the hashtag upon specific violators of ethics outside of Gawker, which still receives fewer mentions than @femfreq.
Additional analysis is needed, but the data could suggest that #GamerGate is capable of discussing general concerns about ethics while remaining incapable of sustaining critical examples of ethical issues within the hashtag.
An inability to maintain a focus on concrete examples of ethical violations while facing concrete accusations of harassment is certainly a fundamental aspect of the current narrative critiquing #GamerGate.
I do plan to look deeper into harassment on both sides, but it will take time as I’m uncomfortable with the reliability and rigor of automated sentiment analysis.
More to come.
