Academic Paper finds Gamergate Supporters Not Significantly Different from Average

An analysis of #GamerGate tweets found that users of the hashtag have older accounts, are more engaged, less joyful, and suspended less often than the average Twitter user. Researchers released their paper titled “Measuring #GamerGate: A Tale of Hate, Sexism, and Bullying” ahead of the Workshop on Computational Methods for CyberSafety in April. While the paper contains few surprises, some of the information is particularly interesting.

Looking at 1.6 million tweets between June and August of 2016, the researchers found that 340,000 unique users participated in the #GamerGate hashtag. Previously, the only data we had about the number of Gamergate supporters was around 150,000 people collected by Chris von Csefalvay back in late 2014 and early 2015.

Further, despite the claims by many that Gamergate is mostly filled with sock puppet accounts created specifically to support the movement, this new analysis has found that Gamergate-supporting accounts tend to be older than a random sampling of baseline accounts. The mean age for Gamergate-supporting accounts was 982.94 days, nearly half a year older than random users, which is 834.39 days.

The report also found that Gamergate-supporting accounts are much less likely to be suspended by Twitter for violations of its rules than the average user. While 20% of the baseline has been suspended, only 9% of Gamergate-supporting accounts have been. The researchers postulate that this may be due to Twitter catching a large number of spam accounts, but present no evidence to support that claim.

The researchers compared the tweets with the Hatebase open source database and found Gamergate supporters use slightly more hateful words than average. The only significant difference in tweet content was that Gamergate supporters tend to be less joyful, they found. “This is particularly interesting because it contradicts the narrative that Gamergaters are posting virulent content out of anger. Instead, GG users are less joyful, and this is a subtle but important difference: they are not necessarily angry, but they are apparently not happy.”

Considering it is a movement of people unhappy with game journalism, censorship, journalism in general, social justice, among many other topics, that makes sense.

The researchers plan to do further research into Gamergate, “focusing on how it evolved over time.”

Update: You can find a live stream of me offering my opinions of the paper here.

  • Smazeli

    Weird, so on a 100-point scale in measuring the hate levels of tweets, the tweets from Gamergate members rated 0.19 points higher? Is that statistically significant?

    Consider that for some reason they lumped in a bunch of other parody and clear hate campaign hashtags in with Gamergate, such as “#InternationalOffendAFeministDay,
    #IStandWithHateSpeech, and #KillAllNiggers”

    It also makes no attempt to discern whether the authors of tweets containing the #Gamergate hashtag are supporters or opponents of the group, which is crucial as a clear number of opponents also used it

    Finally, their citations are also whack:

    “Although held up as a pseudo-political movement by its adherents, there is substantial evidence that Gamergate is more accurately described as an organized campaign of hate and harassment [12]. What started as “mere” denigration of women in the gaming industry,eventually evolved into directed threats of violence, rape, and murder
    [29].”

    [12] https://www.theguardian.com/technology/2014/oct/23/felicia-days-public-details-online-gamergate

    [29] https://www.nytimes.com/2014/10/16/technology/gamergate-women-video-game-threats-anita-sarkeesian.html?_r=0

    Citation 12 contains no evidence of any organized campaign of hate, or any evidence that Felicia was doxed, or that Gamergate facilitated or supported her being doxed in any way. It only quotes debunked allegations from Felicia and Chris Kluwe

  • Thatherton

    The “baseline” is a big problem. Comparing #Gamergate to a random selection of tweets will conflate many variables that we don’t know are Gamergate-specific. It’s possible that any popular hashtag will display properties that are different from completely random tweets. It’s easy to think of ways this might happen (popular hashtags draw in more extreme views, the adversarial nature of the campaign increased over time, etc.).

    The authors should have taken a similarly-sized hashtag campaign (ideally several of them) and used this as the baseline. This would have produced a much more apples to apples comparison where one asks “is #Gamergate significantly more or less X than other similarly sized hashtag campaigns”.

    As it is, we can’t separate #Gamergate-specific attributes from large hashtag campaign-specific attributes.

    • Cornson

      But if they did that they would prove that GamerGate is not a hate movements that wants to drive women out of gaming.

      that’s not the narrative they want.