"Twitter’s focus is on a healthy public conversation."

October 17, 2018 12:51 PM

Twitter has released a dataset consisting of 1.24GB of messages and 296GB of media "that we believe resulted from potentially state-backed information operations on our service", according to the company. They come from "3,841 accounts affiliated with the [Internet Research Agency], originating in Russia, and 770 other accounts, potentially originating in Iran. They include more than 10 million Tweets and more than 2 million images, GIFs, videos, and Periscope broadcasts", and are being made publicly available in partially-anonymized form.

Accounts having fewer than 5,000 followers are included in the dataset with some identifying details replaced with hashes; researchers can obtain the un-redacted versions provided they agree to "a data use agreement that will include provisions to ensure the data is used within appropriate legal and ethical parameters."

The Atlantic Council's Digital Forensics Research Lab, which received access to the dataset prior to public release, has written a series of articles for Medium on what it has called the "Twitter Troll Farm Archives":

1. Part One — Seven key take aways from a comprehensive archive of known Russian and Iranian troll operations
2. Part Two — How the Internet Research Agency regenerated on Twitter after its accounts were suspended
3. Part Three — Assessing an covert Iranian social media influence campaign
4. Part Four — Expanding on key conclusions from the Russian and Iranian troll farms

The fourth article summarizes their conclusions, which include:

1. All Content Points Home
Both troll operations put their governments’ needs first. Russia’s troll operation primarily targeted Russian speakers, while Iran’s focused on pushing regime messaging abroad by promoting aligned websites. The Russian operation’s subsequent use of English-language posting showed how a capability designed for domestic influence could be turned abroad.

2. Multiple Goals
The Russian operation had multiple and evolving goals. One main purpose was to interfere in the U.S. presidential election and prevent Hillary Clinton’s victory, but it was also aimed at dividing polarized online communities in the U.S., unifying support for Russia’s international interests, and breaking down trust in U.S. institutions.

3. Community Targeting
Both operations targeted highly engaged, highly polarized online communities, especially in the United States. The Russian operation attempted to infiltrate and polarize them, while the Iranian operation tried to message them. Any attempts to increase domestic resilience should prioritize working with such communities.

4. Equal-Opportunity Troll Farms
The Russian trolls were non-partisan: they tried to inflame everybody, regardless of race, creed, politics, or sexual orientation. On many occasions, they pushed both sides of divisive issues. It is vital to recognize this factor to end the partisan perception that Russian influence operations focused on one side of the political spectrum. Focus shifted over time or at specific moments based target audience.

5. Opportunism
The Russian trolls often chose targets of opportunity, especially elections and terrorist attacks, in their attempts to interfere in local politics. This included promoting anti-Islam hashtags after the Brussels terror attacks, a pro-Leave hashtag on the day of Britain’s Brexit referendum, and leaks targeting French President Emmanuel Macron before his election. These opportunistic attacks had little to no impact on the target populations.

6. Evolution
Both troll operations evolved, apparently through a process of trial and error in content and messaging. Their activities in 2014 were different from their activities in 2018. Countermeasures will have to take further evolution into account.

7. Low Impact
Other than in the United States, the troll operations do not appear to have had significant influence on public debate. There is no evidence to suggest that they triggered large-scale changes in political behavior, purely on the basis of their social media posts.

Credit for this post is due to MeFi's Own ltl, who originally posted it in the latest Trump/US Politics megathread.

posted by Kadin2048 (16 comments total)

Ay yo, if I was a budding data scientist I would timeline all of this against official GOP talking points and tweets from verified accounts of politicians to track how long it takes for an item to go from Russia>Twitter Swamp>Talking Points>Mike Huckabee.
posted by Damienmce at 1:04 PM on October 17, 2018 [19 favorites]

part two and three above both link to part three
posted by msbutah at 1:06 PM on October 17, 2018

Correct link to Part 2
posted by sysinfo at 1:11 PM on October 17, 2018

My bad. Mods have been contacted. The correct URL for Part 2 is this one.
posted by Kadin2048 at 1:12 PM on October 17, 2018

Damienmce: You'd have to allow for the additional time it takes for Huckabee to punch everything up with his signature wit and whimsy.
posted by Strange Interlude at 1:15 PM on October 17, 2018

Mod note: Link updated!
posted by LobsterMitten (staff) at 1:15 PM on October 17, 2018 [2 favorites]

I'd like to see them post a portal like Facebook did where you could see if you interacted with any of these accounts at any time.
posted by msbutah at 1:16 PM on October 17, 2018 [3 favorites]

That's 1.24GB *compressed*. So probably ten times that, uncompressed. That's a lot of fodder for training a neural network, but I'm going to have to leave it to someone with more bandwidth.
posted by sfenders at 1:19 PM on October 17, 2018 [1 favorite]

I wonder how many of these troll accounts are pushing the "NPC Meme" now.

I'd like to believe it's a forced meme coming from a specific group rather than just lots of people enamored with the idea that people they disagree with aren't real people.
posted by Mr.Encyclopedia at 1:26 PM on October 17, 2018 [2 favorites]

leave it to someone with more bandwidth

Get cloudy. On AWS Lightsail can get 40GB SSD and 2TB transfer for $5 for a month.
posted by Damienmce at 1:27 PM on October 17, 2018

7. Low Impact
Other than in the United States, the troll operations do not appear to have had significant influence on public debate.

The United States being the biggest, wealthiest, strongest, loudest, most interventionist nation on the planet -- I think it's easy to argue that "Low Impact" and "do not appear to have had significant influence on public debate" make for poor, perhaps misleading phrasing.
posted by philip-random at 1:43 PM on October 17, 2018 [2 favorites]

I imagine bandwidth was a reference to time in that case
posted by Slackermagee at 1:46 PM on October 17, 2018

Other than in the United States, the troll operations do not appear to have had significant influence on public debate.

It's very "what have the romans ever done for us?"

Also misleading because there isn't much a look at Russian language trolls effect on the "public debate" in the Russian-speaking world, especially with the conflict in Ukraine.
posted by peeedro at 1:46 PM on October 17, 2018 [2 favorites]

I read the Iranian summary. Not in depth but enough to claim that it would be funny if it weren't so breathtakingly bereft of significance.

- First of all any "troll farm" that attacks the barbarian Saudi regime is in my book on the side of the Angels. I mean do they really think that "Posts on the Kingdom of Saudi Arabia, for example, were routinely hostile and accused the kingdom of terrorism, atrocities, and war crimes" is a false accusation and a bad thing?

- A second point is that I wouldn't trust a NATO and UAE (not to mention Hariri) funded Council, with criminals against humanity on its board of directors, to tell me the time of day, much less what a "healthy public conversation" is. To make these sort of institutions arbiters of what is allowed on social media, is what is truly frightening to me.

- Thirdly they are accused of retweeting mostly *true information* such as "“U.S. has wiped Raqqah off the face of earth.”", "Trump Is Insane". So they are banned for making (and reproducing) valid statements and (in my book again) mostly benign, opinions. This should be contrasted with POTUS being allowed to lie and maliciously slander people through the same platform.

- Fourth, and most importantly, the report actually goes out of its way to not identify this activity as government-run. Which raises the question of "why ban them?", unless they plan to go after all of marketing / political / government accounts [in which case see last point]:

The campaign has not been fully attributed. FireEye tracked it to email addresses and phone numbers in Tehran, but did not identify the organization controlling it. @DFRLab’s own research showed that it consistently shared regime messaging, notably from Ayatollah Khamenei. We refer to it as “Iranian” on the basis of its geography and content; this should not be taken to mean “government run.”

and

While these posts clearly amplified Iranian positions, it is questionable whether they should be viewed as a trolling campaign, in the sense of using social media to engage personally with other users.
The overwhelming bulk of these posts served as advertising, aimed at drawing users towards articles on websites associated with the broader messaging campaign. While they targeted individual users, the posts did not seek to engage with them in meaningful discussion. The purpose appears to have been to draw those users’ attention to pro-Iranian websites.

and most damningly:

One issue on which the Iranian accounts might have been expected to go into overdrive was Trump’s decision to decertify, and then abandon, the Iran nuclear agreement, known as JCPOA. Surprisingly, the Iranian accounts were relatively quiet.

This is, of course, surprising only if one pre-assumes this "troll network" is indeed some state-backed spy / troll network, and not (as seems more likely) Iranian English language publications' marketing campaigns, along with genuine pro-government (remember, the "reformer" - meaning in Westernese: the most pro-western electable candidate - and not the "hardliner" won in the last Iranian elections) local and diaspora views of citizens and citizen- or media- sponsored patriotic propaganda.
- Fifth. It seems that the "Iranian trolls" were actively campaigning against Trump at the same time that the alleged Sinister Tentacled Russian Bot Thing was campaigning against Clinton. Indeed it apparently called them on it in clearly clintonite terms, which is strange given that Russia is Iran's best bet to avoid some US-induced Iraqi-style apocalypse. Those damn trolls can't even coordinate properly. Although Iranian "trolls" seem to be consistently progressive and anti-Trump (a fact that the Idiot in Chief will no doubt use at some point, perhaps as a pretext to invade Iran).
Finally: the 400 pound gorilla in the Middle East government-run bots, trolls and government coordinated social media accounts is Israel (extending beyond the Middle East). Which somehow is never part of any study on coordinated influence campaigns, which are now universally part of western government propaganda. The 10 tonne Elephant, and all time champion in the sport, remains the US, but I'm not holding my breath on a similar "analysis" by the Atlantic Council on the US influence networks on Twitter and FB, anytime soon.
[I really don't look forward to read the Russian story, because making an obscenely huge deal on this is already part and parcel of the liberal American mythology, and this thus needs even less evidence for even more extravagant claims]
posted by talos at 3:18 PM on October 17, 2018 [9 favorites]

I mean do they really think that "Posts on the Kingdom of Saudi Arabia, for example, were routinely hostile and accused the kingdom of terrorism, atrocities, and war crimes" is a false accusation and a bad thing?

Plus, even a cursory glance at virtually anything that's been tweeted on, for example, the Khasoggi assassination in the last week, will quickly reveal that the Troll Farm Gap in the area has already been closed. I mean, I guess it's possible that thousands of different Saudis have individually been moved to post inane, robotic bullshit on every random tweet that impugns the dignity of the Saudi government, but I think there's a more obvious solution.
posted by Copronymus at 4:21 PM on October 17, 2018 [5 favorites]

I'd like to see them post a portal like Facebook did where you could see if you interacted with any of these accounts at any time.


I'll use my business analyst skills to write the requirements for that.

Did user log onto Twitter more than 2 times in 2016? Then yes.
posted by MCMikeNamara at 6:52 PM on October 17, 2018