Thread by @marcowenjones

[#Thread on #Algeria] 1/ @Redasbox got in touch to ask if I could have a look at some unusual tweeting in Algeria. I analysed around 20,000 tweets from around 5,769 unique accounts. What I found, was alarming, and clear evidence of a disinformation campaign. Read on #Algerie

@Redasbox 2/ First some very brief context. Algeria has seen the 32nd weekend of pro-democracy demonstrations. Peaceful protests successfully pressured Abdelaziz Bouteflika, the ruler of 20 years, to step down in April. An election has been called for on December 12th to help try and

3/ resolve the political stalemate in Algeria. However, many Algerians are not happy as they believe the elections will be rigged by a military/elite junta. This is not uncommon in MENA. Thus the elections are seen as giving democratic legitimacy to continued authoritarian rule

4/ Important to bear in mind then that many opposition leaders have called for a boycott of the elections until there are guarantees of an end to elite cronyism. On the other side of the coin, the army are obviously wanting people to go to the polls.

5/ In this regard, two hashtags were mentioned to me as attempting to encourage people to vote #ماتهدرش_باسمي (not in my name) and #الجزائر_تنتخب (Algeria votes/is voting). The 'not in my name' hashtag is criticizing those boycotting the elections, essentially trying to depict

6/ those opposing the elections as being a small, but vocal minority. The hashtags obviously wish to give the idea of support to the military's request to vote. However, as I will show now, much of this support has been 'inflated' by suspicious accounts. #Algeria

7/ So the first most important thing. Most of the 20,000 tweets I analysed were sent between 15th Sept and 28th September. Most tweets in the sample were also sent on 22nd September. See below.

8/ The sample was incredibly skewed. Since 2008, the average number of accounts created per month in the sample was just 44. However, September saw this number rise to an incredible 723 accounts! That's 12%. 12% of accounts in the sample were created this month - since 2008!

9/ That's not all! 474 of those 723 accounts were created in just two days. That's insane, that's around 7% of accounts in the sample were created in just two days. That is usually a surefire way to spot inorganic behaviour and an influence campaign. But what about the accounts?

10/ Well most are designed to look like real people. The attached examples are just a selection of the suspicious accounts. They have names, profile pictures, and tweet in support of the army and the elections.... #Algeria #Algerie

11/ The sample is also strange as the majority are tweets and not retweets, which is unusual in most tweet samples. This means the accounts are trying particularly hard to look like individuals. I can dump some data later so y'all can do reverse image search. Will tweet more in a

12/ bit, but right now I'm gonna eat some chicken.

13/ So yeah, at least 8% of those tweeting for Algerians to go to the elections are probably bots and/or trolls. Stay vigilant. Hopefully @TwitterSafety will suspend them soon. PM if you want the data.

14/ Going to repost this article I wrote recently for @monkeycageblog, as I clearly stated how the recent Twitter takedowns of UAE accounts would not curb influence campaigns. Guess it was true... washingtonpost.com/politics/2019/… #Algerie #الجزائز