This is the second time I've posted this particular tweet. The results were very different the first time - it received 26 replies in a very short span of time, none of which are visible. Best guess is that (possibly automated?) protected accounts reacted to the hashtag.
(previous incarnation of the hashtag-laden bait tweet)
Over the last two days, this tweet in support of Ted Cruz has been tweeted 9635 times by 9014 accounts. Let's take a look.
Very few of the accounts look like bots based on 24 hour activity/use of automation services (and there's a helpful button to make the tweet easy for humans to copy). There's a big caveat though - many of the accounts have too few tweets to perform this test.
Of the 9014 accounts that duplicated the tweet, 908 (9.4%) have default profile pics, and 1610 (16.7%) have fewer than 100 followers. This doesn't prove the accounts are automated, but does suggest the presence of sockpuppets, whether human or automated.
On August 30th, an archive of communication between former US President Bill Clinton and former Russian President Boris Yeltsin was declassified by the White House. We downloaded tweets containing 'Clinton' and 'Yeltsin', resulting in 7200 tweets from 5588 accounts.
Among the themes that pop up in these tweets is the notion that former President Clinton "colluded" to help Yeltsin win, with the implication that any collusion with a foreign power on the part of Trump isn't so bad.
720 accounts have tweets containing "Clinton", "Yeltsin", and either "collusion" or "election" in the last ten days. This chart shows the news sources mostly frequently linked by these accounts. Russian state media seems popular.
This tweet from @shocker_daily was cloned by ten other accounts shortly after being posted. Some kind of automation is the obvious explanation, and the "RT shocker_daily" instead of actual retweets is a telltale sign of IFTTT (TY @MikeH_PR for the lead).
The @shocker_daily account was created recently (7/27/2018) and is almost exclusively focused on posting dubious claims about the #Skripal#Novichok poisoning. It has an associated "news" site (dailyshocker(dot)news), registered the same day with similar content.
The ten accounts that copied @shocker_daily's tweet have been copying its tweets ever since the account started tweeting. The cloned tweets are, as suspected, posted via IFTTT. The majority of @shocker_daily's reach is via the bots, as @shocker_daily has very few followers.
If you've been reading the analysis threads @ZellaQuixote and I post, you've probably seen the tweet schedule plots we include. Here's a simple web application you can use to make those plots yourself for accounts of your own choosing. makeadverbsgreatagain.org/allegedly/
The plots include 6-7 weeks of history ending with the most recent tweet, limited to the most recent 3200 tweets. If 3200 tweets don't go back the full 6-7 weeks, part of the plot will be blank. The circles are color-coded based on the app used to post the tweets.
This tool has been updated to include a more comprehensive list of timezones.
On September 1, 2018, one or more explosions occurred at Mezzeh airbase in Damascus. There are competing stories as to the cause. We downloaded tweets containing "Mezzeh", resulting in 4764 tweets from 3588 accounts, 246 of which (6.9%) were flagged as likely to be automated.
When testing accounts for automation, a metric @ZellaQuixote and I have been using is whether the accounts A) post 24/7 and/or B) post 90% or more of their tweets via automation apps/services. We decided to apply this test to our own followers.
Of 32045 accounts that follow one or both of us, 232 tested positive for automation per this test. This chart shows the 25 most common apps/services used to post the tweets. IFTTT and Buffer are the most prominent.
There are at least 645 accounts in the set. All of them follow the same naming scheme: [female name] + [2 digit number] + [letter] + [2 digit number] + [letter]. The vast majority of the profile pics are female, presumably stolen from elsewhere on the Internet.
These accounts are old - created within a 38 minute span on August 6th, 2012. This set of accounts was created long before they became highly active, and not all are active yet - of 645 accounts, only 95 have tweeted.
Two days ago (8/24) a new account (@u_franken) theoretically belonging to former Senator Al Franken was created. Both @ZellaQuixote and I are old enough to remember when it was totally cool to say "inbox me" to random strangers, which is to say never.
In addition to telling people to "inbox me", the fake Al Franken account @u_franken also posts copies of the real Al Franken's tweets from his @SenFranken account. If you're looking for Franken's real post-Senate account, he can be found at @AlFranken, btw.
An interesting story about security and human nature. TLDR: for the first decade or so of the McDonald's Monopoly contest, it was utterly impossible for the average customer to win any of the top prizes. thedailybeast.com/how-an-ex-cop-…
The dude hired to oversee the safe and secure handling of the super high-dollar prize pieces (Boardwalk etc) absconded with them. He started off by offering a prize to a friend in exchange for a cut of the money.
Once that worked out he got more methodical - looked for folks in a variety of states so the geography of the winnings would be randomly distributed.
Both @ZellaQuixote and I just added "make a bot that might outlive me" to our bucket lists. On that note, let's take a look at a bot!
Meet @Cary88888888. This account has been around for quite a long time, has 18.9K followers, and has tweeted more than 2.6 million times. There are some early organic tweets, but the vast majority are posted via automation services (breakdown for the last two years shown here).
Here's @Cary88888888's tweet schedule at two different time periods. In both cases, the majority of the tweets are posted via IFTTT. It's likely that the change in schedule around the beginning of the year is due to a change in IFTTT's behavior (we've seen it on a few accounts).
In the aftermath of her suspension from Twitter, @caitoz and others have made the claim that she was attacked by a "bot army" and that this action was the cause of her suspension. Let's take a look.
This is the tweet that was theoretically amplified and reported by a bot army. Let's test the accounts that retweeted, quote-tweeted, or copied this tweet for signs of automation (24/7 tweet schedules and/or use of API/automation services).
Of the 775 accounts that retweeted and 53 that quote-tweeted @TrickFreee's tweet, exactly one was flagged as a likely bot based on the test described in the previous tweet.
What does @Ohio_Buckeye_US tweet about? 41.4% of the account's tweets contain one or more of the terms shown in the chart. This political independent dutifully checks the major #MAGA propaganda boxes. Tweet schedule is high volume and bursty.
Let's take a look at the hashtag #DissolveTheUnion, one of several associated with a proposed Scottish independence vote. There have been 25316 tweets from 3476 accounts containing the hashtag in the last 10 days.
Only 31 of the 3476 accounts with #DissolveTheUnion tweets appear to be automated (based on 24/7 tweeting and/or use of automation services). However, they are responsible for 3925 of 25316 tweets (15.5%). The majority of the volume is coming from the top three accounts.
Generally charts and graphs are a good way to visualize data, but it's fun to play with other mediums sometimes. I first tried experimenting with audio by converting the tweets of (now-banned) Holocaust denial spambot @RealFKNNews into a drumbeat.
Just for fun, let's try the same thing with pro-Brexit troll DavidJoBrexit (formerly DavidJo52951945). The tweet timings are very orderly, which makes converting to rhythm easy.
I decided to drop the dead space and use each hour as an eighth note, resulting in 14/8 time. Melody was determined by the offset of various tweets from the hour boundary (and then tweaked somewhat in service of it sounding better).
This is an interesting idea. Here's a quick proposal for how to implement it.
Automated tweets would be marked with a robot icon. The marker is at the tweet level rather than the account level because some accounts mix automated and organic tweets.
Some bots are spammier than others, and unless one searches through the user's tweets this isn't always obvious. Repeated automated tweets would thus be marked with the number of times the tweet in question had been posted alongside the robot icon.
Hypothesis: individuals tweeting organically about US politics will (consciously or subconsciously) favor material related to states to which they have ties. Bots and managed accounts, on the other hand, will follow news cycles and upcoming elections.
To test this, we downloaded tweets containing the names of 14 states from the week before each state's primary. We filtered this set to accounts that had tweets for at least 3 states and one or more of a set of political terms/hashtags (#MAGA, #Resist, etc) in their profiles.
This yielded a total of 790 accounts. Checking the set for probable bots (24/7 activity and/or 90% of tweets posted via automation tools) yielded 38 accounts. 3 were false positives; the rest we categorized as right-wing, left-wing, news or humor.
Meet @Evelyn_Fordd. This is a pro-Trump account that is automated via three different services: dlvrit, twitterfeed (which appears to have shut down), and jurijkovall (we'll return to that one later).
Searching through the accounts that have interacted with @Evelyn_Fordd turns up two additional accounts using the same set of automation services: @Doris_Finchh and @Carol_Bushh. Note the weird repeated letter in the last names of all three.
We then downloaded tweets for @Evelyn_Fordd followers and followees (is that a word?) and found a total of 20 automated accounts fitting the same pattern with a variety of tweet schedules. They appear to have been created in two batches: one in July 2016, the other in April 2017.
Over the last several days, a new hashtag has appeared and propagated on Twitter: #VerifiedHate. The concept seems to be to attack verified (blue-check) accounts that are theoretically promoting hate against white people.
We downloaded recent tweets containing the #VerifiedHate hashtag. We found 18092 tweets from 9747 accounts - the first tweet is in the wee hours of the morning on August 5th, and the hashtag takes off on the evening of August 7th.
This hashtag campaign appears to have been planned on other platforms in two stages. First, we have this post on Gab from 8/5, which is within minutes of the first tweet from @Keque_Magus. It's followed by plenty of additional Gab discussion promoting #VerifiedHate.
AMERIA FIRST!!!! Typos - the gift that keeps on giving. #AmeriaFirst
Here's an obvious bot: @MagaNewsB. Its recent tweets have an amusing typo: #AmeriaFirst rather than #AmericaFirst. A potential pitfall of automation is that if you accidentally configure a bot to tweet a typo, it'll tweet the same typo over and over, and the typo may proliferate.
The first use of the hashtag typo (#AmeriaFirst) is this tweet from @FoundFathersC on 8/6. Subsequently, it spread to 14 other accounts - and the first four uses were within five minutes.
If #QAnon is a prank, it is a reckless and dangerous one that's not unlikely to inspire real-world violence in the current political climate (see #Pizzagate shooting.) I don't know who is behind it, but it smells more like a psyop/propaganda campaign than an attempt at humor.
To clarify - my main issue with this article is the "extremely likely" aspect. The evidence simply isn't there to back that assertion (the only data point in the article is the book parallel), and there are a number of other possible origin stories for #QAnon.
The theory presented in the article also ignores the provided explanation for the name "Q" (a federal employee with Q clearance, see link), as well as omitting the continuity between previous related conspiracy theories (i.e. #Pizzagate) and #QAnon. en.wikipedia.org/wiki/Q_clearan…
We decided to compare accounts based on their perspective on the #QAnon conspiracy theory. We tried a few different techniques for classifying accounts - the one that worked best involved analysis of all tweets containing the phrase "Qanon is" from the last five days.
We used the retweet network(s) for the "QAnon is" tweets to classify the accounts. The graph separates cleanly into three sections: #QAnon true believers, #Resistance skeptics, and #MAGA skeptics. We manually inspected the most prolific accounts to verify the classification.
The true believers have, compared to the other 2 groups: (1) a higher frequency of accounts with default profile pics, and (2) a more even follower/following ratio, suggesting followback accounts.
Meet @VictorialovesAm, an account with a default profile picture and almost 20K tweets in a little over four months.
Here's @VictorialovesAm's tweet schedule. The majority of the tweets are concentrated into brief, spammy tweetstorms, with isolated instances of more organic activity. Let's take a closer look at the burst of activity around 11-noon on July 30th.
In the span of 50 minutes on 7/30, @VictorialovesAm tweeted 152 times (mostly RTs). Several accounts were retweeted 3 or more times in that vicinity, but interestingly retweets of the same account are spread across the time range rather than being back to back.