Thread by @veikkoeranti

What is politics in online anonymous imageboards (think 4chan, but in Finnish)?

We have a 🚨new paper🚨 out, answering just that! Here's a #thread unwrapping computational & qualitative investigation of Ylilauta done with @tuukkaya & @SKRHardwick

1/x

The paper is just out at @SocialMedia_Soc, and is available in glorious #openaccess at journals.sagepub.com/doi/10.1177/20…

2/

Ylilauta (Overboard) is a Finnish translation of 4chan – it runs the same basic software, has the same anime, and the same naz.... alt-right tendencies. But there's more to it!

We wanted to find out what is the big picture of politics on the board, and boy was that a ride!

3/

Why Ylilauta? B/c its one of the places where political culture is created, where memes are conjured into existence, and where esp. young men practice politics, in both senses of the word.

4/

One of the neat things about computational social science is that you don't have to impose your own conceptions into data (at least as much). We wanted to know what's political for the community at Ylilauta, and how much of Ylilauta is political talk.

5/

We had a large chunk of Ylilauta available for analysis, about 800k messages. We started from the politics sub board, and @SKRHardwick created word embeddings / convolutional neural network thingy for us which learned the features (vocabulary etc) typical for political talk

6/

(For the methods, if you read Finnish, check this thread!

https://twitter.com/SKRHardwick/status/1330156283492786181

)

Using this neural wizardry, we scored all the messages in our corpus and did a bunch of involved two-person qualitative validation to be sure that our model understood politics in a way that was also comprehensible for humans (not a given, always validate qualitatively!)

7/

All in all, about 10% of the whole Ylilauta was political talk in our sample. So a significant chunk, but still only a small part of the whole. We found political talk in all the sub boards, with immigration and science boards trailing /pol/ in quantity.

8/

And what was that political talk about? We used a LDA topic model to group all the political talk into 10 topics, ranging from surprisingly deep excursions into political philosophy to outright racism.

9/

One benefit of our inductive definition of politics is that sexual assault clearly is a political topic for Ylilauta (and not in a feminist sense, more in connection with immigration), and I am not sure we would have categorized it that way in trad qual analysis.

10/

We actually had two separate chunks of Ylilauta, collected in 2014-2015 and 2018-2019, so we could look at how the popularity of these topics developed.

Turns out Ylilauta turned pretty clearly towards far-right/alt-right during that time period.

11/

Discussions abt race, sexual assault and immigration were dominating the political discussion, and themes of philosophy and economics were disappearing. This of course coincides with the whole maga-palooza, as well as the mainstreaming of these themes in Finland.

12/

We also did a qualitative reading of a subset of the most political messages. Here we utilized the idea of group styles by @NinaEliasoph – how the users delineate between themselves and various others.

13/

The whole tone on Ylilauta, as also in 4chan, is irreverent, mock-everything, everything is ironic and I am a cool edge-lord above caring.

The main division is between 'us' and the 'normies' – those in the know of the chan culture and norms, and those outside of these.

14/

This division was not neutral – those delineated outside were also categorized as 'hippies', 'sjw', 'feminazis' etc (perhaps not surprising).

So even with this everything-is-ironic blasé attitude, the group boundaries were clearly used to enforce (alt-)right-wing politics.

15/

Tl;dr: Ylilauta is plenty political, but politics is only a minority subject there. Far-right themes increasingly dominate, but you can also find deep political philosophy etc. The overall tone is ironic but enforces a right-wing identity.

16/

Takeaway: this is important for political analysis – these spaces are not only 'seedbeds for radicalization' or what have you. When that sort of thing happens, it happens alongside anime flamewars, confused teenage sexualities etc. Understanding whole community is crucial.

17/

Getting long but still some caveats: what did we NOT see?

Since our whole analysis is based on the boards own definition of politics, there might be (even a lot of) stuff that for a sociologist would have clearly been political if surfaced.

18/

One clear omissions are the other gendered themes except sexual assault: many analyses of similar web fora have shown rampant misogyny and the whole #incel stuff. In terms of text style it is clearly being discussed in different register than what's 'politics' for them...

19/

...and one a similar note, we know that these board orchestrate harassment campaigns, comment raids, spam, doxxing and so on. These are also not 'politics' for the users, and thus not visible in our analysis. Our analysis is skewed towards more analytical talk.

20/

And in that analytical talk on Ylilauta, racism is political but feminism is not, or at least it's not as big a theme on the board.

21/

And another caveat: our analysis was strictly text only! We were actually quite surprised how much sense everything makes without the images, even though the thing's called an imageboard.

Doing a multimodal analysis would be really cool!

22/

And that's a wrap! The article is #openaccess so you can also see all the juicy details yourself. Thanks @tuukkaya and @SKRHardwick, it was a pleasure!

journals.sagepub.com/doi/10.1177/20…

/fin

Addendum: the paper is a joint effort but all the snark is purely mine, Tuukka and Sam would have different snarks, different emphasis, and probably different politics as well :)