Quick update - I ran a bit of analysis w/ @antmandan of these suspicious accounts, by collecting all tweets since 1/1/2022 that mention @\sydney_festival and analysing this network of coordinated behaviour
The network above shows coordinated behaviour where accounts (nodes) are connected if they both performed the same action within 60 seconds of each other, at least twice.
We focus on the big cluster of coordinated reply activity (red links) that got folks' attention
2/9
These are the accounts that were posting positive replies to @\sydney_festival tweets. They're clearly part of a coordinated network that spams copypasta replies mainly about crypto, but also various companies, events and personalities
3/9
They don't really hide the fact that this is a paid PR astroturfing operation. And they tweet in multiple languages including English, Arabic, Turkish and Chinese
4/9
It is perhaps for this reason that they score highly as bots on the 'Language-independent' @Botometer metric, but aren't particularly bot-like based on language dependent features
5/9
I don't think they're bots - at least not fully automated accounts - but they probably use some kind of scheduling software and are controlled by a central agent, i.e., semi-automated sockpuppets for hire. Their timing and volume of tweets is almost but not quite identical
6/9
So did @\sydney_festival hire this paid PR operation? The answer is: we just don't know (and almost certainly can't know based on Twitter data alone). So we can only speculate at this point
7/9
Three hypotheses: 1. They were hired by @\sydney_festival or some entity trying to improve the brand reputation 2. They randomly replied as part of common practices used by infops to cover tracks and look more organic 3. Someone else hired them and wanted it to be found
8/9
Hope this thread was helpful to shed some light on this interesting twist in the Sydney Festival fail.
If anyone would like any details of data collection, network construction, tweet IDs, etc, please feel free to DM
9/9
• • •
Missing some Tweet in this thread? You can try to
force a refresh
But when it comes to retweets, Kelly's metrics are overshadowed by the amplification efforts of a coordinated network of fringe protestor accounts, as this chart shows:
We see that Kelly doesn't appear in the coordinated behaviour network (see original tweet above). Rather, his tweets frame the agenda and injects a massive amount of fuel into the Twitter protest networks *but* he doesn't coordinate his activity with them.
Folks messaged me about a new hashtag anti-Andrews campaign trending, so I collected over 67,000 tweets (past 7 days) containing the relevant hashtags and ran the data through a coordinated behaviour detection system
A few insights, starting with a network map 🧵
The network shows large-scale, loosely coordinated behaviour by a hard core of campaigners who spam the hashtags repeatedly together, all day long.
BUT it's @CraigKellyMP and @OzraeliAvi who jump onto the fringe activity and turn it into a roaring blaze with two viral tweets
A pro-Andrews counter-attack started almost immediately, orchestrated by @PRGuy17, in an effort to hijack the hashtags and drown out the anti-Andrews discourse
New study reveals private groups behind the 'pl*ndemic’ disinformation campaign. These groups not only coordinated a surge in COVID-19 conspiracy theories, but also "coached" citizens into fanatic activism against COVID-19 measures. frontiersin.org/articles/10.33…
The study found that this was driven mostly by small-reach conspiracy theorists all posting at once:
“The most influential Twitter users … appear to be either citizens or activist accounts, rather than bots”
Also interesting:
“And the most common word in more than half of all top users' profile descriptions was “truth.” These profile descriptions often signal a search for a “hidden truth,” as if they are part of a citizen initiative to purge the world of evil actors.”
Just to clarify, the primary interest here is Twitter's design features and dynamics, and how a single account (@PRGuy17) managed to get a hashtag to #1 on the Australia trending list in less than 1 hour
I ran sentiment analysis to provide a quick comparison with our peer-reviewed study of hashtag publics during #Covid19Vic, where we found that pro-Andrews tweeters were overwhelmingly positive and anti-Andrews tweeters were overwhelmingly negative: