Read on Twitter
Here's a thread on this graph I made, which has been bouncing around Twitter a bit. Broadly, it shows the connections between "influential" Twitter accounts during the election.
But it's always more complicated than that with networks! Here's some insights you might be missing.
But it's always more complicated than that with networks! Here's some insights you might be missing.
First, some details on how it was made:
• Data collected from Sept. 1 - Dec. 1 on election-related terms ('voting', 'dominion', etc.) by @uwcip
• Each node is an account and a link between nodes means that both accounts were retweeted multiple times by the same group of users.
• Data collected from Sept. 1 - Dec. 1 on election-related terms ('voting', 'dominion', etc.) by @uwcip
• Each node is an account and a link between nodes means that both accounts were retweeted multiple times by the same group of users.
• Nodes are drawn closer together if relatively more users retweet both accounts.
• Nodes are sized according to how many links they have to other nodes.
• Nodes are colored according to a clustering algorithm in open-source software package @Gephi.
• Nodes are sized according to how many links they have to other nodes.
• Nodes are colored according to a clustering algorithm in open-source software package @Gephi.
• Black nodes have been suspended or deleted as of Jan. 20, 2021. Protected accounts are not black.
That part's important: these (potentially) aren't just QAnon removals. A few accounts have been reactivated since that date, and could have been removed for other reasons.
That part's important: these (potentially) aren't just QAnon removals. A few accounts have been reactivated since that date, and could have been removed for other reasons.
But given the subject matter and timing, many are likely QAnon removals. Which brings us to the clusters. Let's start at the bottom: right-leaning accounts.
We've struggled on how to accurately label these accounts, so I'll instead just describe.
We've struggled on how to accurately label these accounts, so I'll instead just describe.
The yellow cluster is populated with accounts that incessantly retweet and sometimes tweet pro-Trump content, and sometimes QAnon content too. They almost always have no non-Twitter presence.
Most fit the profile described by @3r1nG in this article: https://erin-gallagher.medium.com/trump-trains-84bea1c3170d
Most fit the profile described by @3r1nG in this article: https://erin-gallagher.medium.com/trump-trains-84bea1c3170d
The red cluster is mainly right-wing accounts with more regular Twitter behavior. They produce more original content, and often are not anonymous or have non-Twitter presences. These include politicians, famous Trump lawyers, talk show hosts, and generic far-right influencers.
Importantly, some of the biggest nodes in the red cluster are ultra-popular QAnon influencers. According to this metric, their election retweet audience is very similar to that of a less conspiracy-oriented right-wing influencer.
This is an important point: network viz's can show people what they want to see. There's a big narrative that QAnon is a "cancer" that can be excised from the right, and this graph appears to support that. But truly, QAnon is central both in the red and yellow clusters.
The yellow cluster has different behavior from the red, though. In our work at the @2020Partnership, we often found that it exploded with election misinformation slightly before the red, for example. Probably because members of the yellow cluster retweet pretty much anything tbh.
An earlier version of the graph (with animation!) and related analyses on election mail-dumping narratives can be found in this post from @uwcip and the @2020Partnership: https://www.eipartnership.net/rapid-response/mail-dumping
Meanwhile, we have the left-leaning accounts in shades of blue. The main blue cluster are journalists, left-leaning politicians, and other influencers.
The dark blue cluster is the left equivalent of the yellow cluster: anonymous and frequent users with large followings.
The dark blue cluster is the left equivalent of the yellow cluster: anonymous and frequent users with large followings.
The light-blue cluster is relatively more left-leaning accounts, mostly focused on Bernie Sanders and closely-aligned political movement.
Some of these accounts were removed, but remember: it could be for anything. It seems unlikely that these accounts were suspended for QAnon.
Some of these accounts were removed, but remember: it could be for anything. It seems unlikely that these accounts were suspended for QAnon.
@informor's group ( @AntonAbilov, @yiqqqing, @hanamatatov) has been doing a lot more work on the suspensions, which you can check out! https://twitter.com/informor/status/1352289036854046725
Anyway, here's some FAQs I've been getting about this graph. Number 1: Where's Donald Trump?
The biggest black node of course, straddling the border between red and yellow.
The biggest black node of course, straddling the border between red and yellow.
Which accounts are between the top and bottom clusters? Mostly, news networks calling election results and vote tallies.
Online news site the Hill is frequently placed between the two clusters in these networks. Visit their site and you'll get it: headlines for both sides.
Online news site the Hill is frequently placed between the two clusters in these networks. Visit their site and you'll get it: headlines for both sides.
What's up with all those tiny nodes below Trump? Trump
uniquely attracts attention from many communities otherwise removed from the right-wing core.
One example: one tiny node that only connects to Trump is an account that translates all of Trump's tweets to Japanese.
uniquely attracts attention from many communities otherwise removed from the right-wing core.
One example: one tiny node that only connects to Trump is an account that translates all of Trump's tweets to Japanese.
That's it for now :).
Last point: look at graph's like this critically! It's easy to get carried away by something shiny, and I encourage you to look and criticize away about what this may be misrepresenting, or excluding. I'm doing so myself right now, for next time..
Last point: look at graph's like this critically! It's easy to get carried away by something shiny, and I encourage you to look and criticize away about what this may be misrepresenting, or excluding. I'm doing so myself right now, for next time..
You can also see this graph shared in this quote tweet from @katestarbird ... https://twitter.com/katestarbird/status/1358088750765539329
and in this Nature News story! https://www.nature.com/articles/d41586-021-00257-y
