“Many accounts on Twitter are created for propaganda”

Rizwan Saeed, an online big data researcher, talks to Soch about his frustration with ‘propaganda fake accounts’ and the existence of political party-backed social media networks of trolls.

Rizwan Saeed, a cultural anthropologist, is currently completing a Master’s degree in Social Research from Australian National University in Canberra. Saeed’s work lies at the intersection of gender and social network analysis and this makes Pakistani Twitter a ripe research ground for him. Recently, Saeed shared a few graphs on Twitter; some explained how the trolling of Pakistani journalists (in late April) was an orchestrated event, another displayed how there was an organized twitter campaign to support the presidential system over the parliamentary system, and a third showed the organic digital spread of Aurat March.

In a conversation with Soch, Saeed talks about his frustration with the misleading phrase ‘fake accounts’; the existence of political party-backed social media networks that have the capability to harass a variety of Twitter users, and generate trends; and the gendered nature of trolling. Excerpts of the conversation follow:

Maham Javaid (MJ): Recently, there’s been a lot of talk about ‘fake’ accounts and ‘paid’ accounts. What’s the difference between them?

Rizwan Saeed (RS): In my research I avoid the phrase ‘fake’ accounts. When you use the word ‘fake’, it implies there isn’t a real person behind the screen. ‘Fake’ accounts have become a blanket term for accounts that don’t appear to be genuine. There are many accounts, call them fake or whatever you like, that are created for propaganda.

Based on my experience of analysing Twitter data, I have categorised Twitter accounts into five main categories: First we have the ‘Verified’ accounts, those acknowledged by Twitter’s famous blue tick. Next, we have accounts that are ‘Genuine but unverified’. Examples for this would be @ZarrarKhuhro or @XadeeJournalist. Then we have ‘Identity-based propaganda accounts’ – these are accounts where users reveal their identity and extensively tweet at and troll particular parties or individuals. They maintain online networks. An example would be @faisalchattha1. After that we have ‘Veiled genuine’ accounts. These accounts seem genuine based on their twitter activity but they don’t reveal names, genders or locations. Within Pakistan’s cultural context, sometimes people can have genuine reasons for keeping their identities hidden. Such accounts aren’t trolling anyone. And lastly we have ‘Veiled propaganda accounts’. These accounts have suspicious identities; they might be using female names or pictures. Examples for this would include @Troll_Keeper or @farinaikram.

An example of a veiled propaganda account

MJ: How can Twitter users tell if an account is genuine? Can you spot a propaganda account that’s being paid?

RS: There are a few identifiers you can observe to tell if an account is genuine. Firstly, take a look at the Twitter handle: are they revealing an actual name, a photo, a gender? Your first clue will be opaque handles like @IlovePakistan3901 or @SA8547. In my observations, these are people who repeatedly make new accounts, and grow tired of thinking of new names, especially since their identities don’t matter, so they let Twitter suggest names for them when they are setting up their accounts. That’s where the strange assortment of numbers is coming from.

Next, do a simple comparison of when the account was created and how many time the user has tweeted. If an account was made a month ago and has already tweeted 3,000 times, that’s another hint that they are using Twitter for propaganda. A genuine Twitter user wouldn’t tweet that much. Thirdly, look at their timeline, if all they do is troll specific groups or individuals, you likely have found a propaganda account.

There are many accounts out there that are created for propaganda, now whether they being paid or not is difficult to tell. Sometimes the people operating ‘paid’ accounts will admit that they are being paid for their tweeting labour, other times they will not. Who is to say? But regardless of the payment, they can’t be considered as genuine Twitter users.

If you look at Pakistan’s Twitter data between Feb-April this year, the most discussed and debated event was Aurat March.

MJ: So, are political parties only use real humans for propaganda and trending hashtags on Twitter, or are they also dabbling with bots?

RS: In my observation, the Pakistani twitter sphere has very few bots. This is quite believable because labour here is so cheap, why would you need bots? I know a few cases where various individuals and even groups are offering their services for propaganda on Twitter. I also know of political parties that pay young adults roughly Rs5,000 – Rs10,000 per month for fulfilling a daily quota of 400-600 retweets a day.

There are also cases of young boys, inspired by social media activists who made their way to mainstream media through enough viral posts, searching for quick fame through Twitter. Many young kids are looking to be gain notoriety on Twitter. At times, they aren’t even being paid but they are doing propaganda for a certain party for cheap fame. I have been approached by young kids to put their Twitter handles on my graphs so they can harness some public attention. There is no dearth of willing human labour in Pakistan.

MJ: In a news report published in April this year you said that one way to decipher how genuine a tweet or hashtag is to see what kind of engagement it receives. What did you mean by this?

RS: There are handful of ways for someone to engage on Twitter: they can tweet, retweet, mention, reply mention, reply and quote someone. But often a tweet will just have hundreds of thousands of retweets and not much else and that leaves you wondering why the people retweeting it aren’t engaging in debate?

You have to look at it on a case to case basis. If there’s been a bomb blast and there’s a post sharing information about the blast; or its Jummah and there’s a beautiful picture with ‘Jummah Mubarak’ written on it; there’s no need for debate. Retweets of such tweets would be organic.

But if we consider the hashtag about the promotion of presidential form of government, that’s an opinion that calls for debate or discussion. That happens when you post something and someone else replies, or if you retweet me, you do it with a comment on top. But there was not much engagement on that hashtag other than retweeting. A particular group was retweeting that hashtag to ensure that it became a top trend.

MJ: What’s a genuine hashtag that you’ve observed on Pakistani Twitter?

RS: If you look at Pakistan’s Twitter data between Feb and April 2019, the most discussed and debated event was #AuratMarch. To set up a comparison for you, the retweet percentage around Aurat March was 64 per cent; the rest of the engagements comprised 11 per cent replies, and 18 per cent of reply mentions. This can be described as a healthy mix of engagement.

Compare that to the retweet ratio of another viral hashtag: #KaptanKiNakammiTeam which stands at 96 per cent. A Tehreek-e-Labbaik Ya Rasool hashtag I have been following has a 97 per cent retweet ratio.

MJ: On and off, when you present analysis on Twitter, you mention that certain groups may be “toeing PTI’s social media strategy”. Could you clarify what you mean by this?

RS: I think that’s quite obvious. Pakistan Tehreek-e-Insaf’s (PTI) strategy includes creating network groups that are headed by group leaders. Within these networks there are certain accounts that can’t be considered as genuine accounts, they exist solely for propaganda and to promote the party’s ideology. Tweets from these accounts, with their opaque handles and veiled identities, are retweeted in the thousands, especially with network groups.

Other political parties are far behind than PTI in these trends. Pakistan Muslim League-Nawaz (PML-N) is trying to expand its social media cell, they also have network groups, but data shows that they are not too organised just yet. Often, Jamiat Ulema-e-Islam and PMLN have joined hands to trend a hashtag against the government or PTI.

Various political parties’ social media strategy includes creating network groups that are headed by group leaders. Within these networks there are certain accounts that can’t be considered as genuine accounts, they exist solely for propaganda and to promote the party’s ideology.

MJ: Let’s talk about what a network group is.

RS: Certainly. Graph 1 is a representation of a network group. The circles, called nodes, are Twitter accounts; the lines represent engagement; pink is for retweets; blue is for reply mentions and green is for mentions; the thickness of each line depicts the volume of engagement between two nodes.

Graph 1

This particular network group is an ‘in degree centrality network’, created to troll journalists (@XadeeJournalist, @MarviSirmed and @UmarCheema1). The network group shows the central players in this particular hashtag and how active each of them were. It’s depicting in bound ties between the central players. In bound ties means that the people in this network are either being abused/trolled, or they are abusing/trolling through engagements, which in this case are retweets. When A retweets B’s tweet, B would get an inbound tie (arrow head) and A will get an outbound tie (arrow tail). In this case the journalists are being targeted and the trolls are retweeting each other.

On the right hand side, you can see four big nodes, who in this case are trolls: @IKarachiwala, @Rana_OkaraPTI, @Troll_Keeper, and @Lalika79. These are the people who hold key positions in this network: they tweeted themselves and they were also heavily retweeted. For example, @Azhar_TanolY heavily retweeted @Rana_OkaraPTI, you can tell this by the pink line, the arrow pointing towards Rana, and the thickness of the line. Similarly, @Rana_OkaraPTI has retweeted @IKarachiwala, and has tweeted a huge amount of reply mentions about @Troll_Keeper.

Most of the people in the network, excluding the journalists of course, are allies of these trolls, but there are also some victims in the mix; for instance, @BBhuttoZardari and @AbdullahDayon are victims of trolling in this network.

MJ: And what can network groups tell us about the way trolls, paid or otherwise, operate on Twitter?

RS: Let’s take a look at Graph 2 to understand this better. In Graph 2, you can see @Lalika79’s group from within the larger network group. And @Lalika79 is the group leader. Firstly, if you look at each of the profiles in this group, you can see, based on our earlier definitions that they appear to be propaganda accounts or accounts that aren’t genuine.

Graph 2

Now look at Graph 3 and 4. You can see that the smaller trolls, such as @Gumsum4, in each leader’s group are the same, i.e. the same people are engaging with @Troll_Keeper, @Lalika79 and @Rana_Okara PTI. This shows that there is a structured network present.

Graph 3

The second thing these networks reveal is lead figures, or leaders. The purpose for graphs 3 and 4 is the same: to identify the networks and the communication that’s taking place within them.

Graph 4

MJ: You took the time to make a special graph (Graph 5) for journalist and activist Marvi Sirmed. Why was that?

RS: Yes. In my research, I have noticed that women are targeted more than men by these trolls. In this instance, the trolls began a campaign against the five journalists in Graph 5, but the tweets regarding Marvi Sirmed were mentioned, quoted, and retweeted more than that tweets about any of the men. Simply put, it’s obvious that Marvi Sirmed was targeted more than others on the list.

Graph 5

MJ: I see Aamir Liaqat is also standing out in this graph, was he also trolling the journalists?

RS: Actually, Aamir Liaquat is marked in blue, because his account didn’t behave exactly the way the troll accounts did, but it’s still part of the network group. He tweeted twice, using derogatory hashtags, and since he is a public figure, his tweets were retweeted a lot. That’s why he became part of the network.

MJ: Right, to go back to the topic of women being more targeted on social media, what did you learn from your analysis of Aurat March tweets?

RS: The two graphs I shared about Aurat March are 1.5 degree Ego Networks (Graph 6 and 7). The main topic or handle is at the centre and the distribution shows us who on Twitter is talking about this topic. To keep it simple, the pink lines are women and the blue are men. In the pink cluster, women are talking to women, in the blue cluster, it’s men conversing with men, and where you see pink and blue overlapping, men and women were conversing. There were more women than men tweeting about Aurat March – which is interesting given that the ratio of men to women on Pakistani Twitter is 70:30 – and the majority of them were supportive of the march.

Graph 6

The most stunning takeaway from this graph is that it shows that the Aurat March network was the biggest 1.5 degree ego network for any node, on Pakistani twitter this year, bigger even than the networks espousing state narratives.

MJ: So would you say that the conversation around Aurat March was mostly conducted by genuine Twitter users?

RS: To answer that, we should look at Graph 7. The purpose of this ego network is meant to show online bullying and harassment regarding Aurat March. The green nodes are women; purple are men; grey are media groups or other companies. These seem genuine. Now have a look at the and orange and pink nodes. These are accounts that deactivated themselves or were suspended by Twitter within 20 days of the March.

Graph 7

Now, notice the width of the orange and pink lines. Notice how thick the orange lines are, this shows us, they created these accounts, abused and trolled Aurat March as much as they could, and then deactivated their accounts. Their entire purpose of existence was to troll.

So yes, the majority of the accounts tweeting about Aurat March were genuine but the trolls also fully participated, before disappearing forever.

– This interview was edited for length and clarity.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *