Tagged: hate speech

Experimenting in Mapping Online Anti-Shia Sectarianism on Twitter in the Middle East

ballsWouldn’t it be interesting to see if sectarianism itself was more dominant in one place than an other, at least online? Are some countries/cities more sectarian than others? Is sectarianism a localised phenomenon, despite what we might see in the news? If we knew this, we could then highlight where to prioritize tackling it.


In  order to do this, I conducted an preliminary experiment. Firstly, I would require some way of trying to determine where a piece of sectarian discourse came from. I decided to locate sectarian tweets, as often Twitter accounts come with information about location.

I thought I’d approach ‘anti-Shia’ sectarianism, as the terminology is familiar, and more prevalent (See Alexandra Seigel’s work). I will be doing the same with ‘anti-Sunni’ Tweets too. (I should add, for the record, that I find the terms somewhat grotesque, as the nature of sectarianism cannot be reduced to a binary). Hopefully though, locating the geographic prevalence of specific discourses challenges the problems of essentialising sectarianism as a monolithic and ubiquitous muslim-wide global issue.

Anyway, to test/experiment with this method, I extracted approximately 10,000 tweets that ranged over a four day period (8th – 12th May 2017). These tweets contained at least one of the following, commonly used derogatory terms referring to Shia.

Twitter Search Rule: “ابناء المتعة” OR “روافض” OR “رافضة” OR “اولاد المتعة” OR “مجوسي” OR “المجوسي” OR “صفوي”

The terms largely relate to religious-sectarianism, such as Rafida, Awlad/Ibna al-Muta’, Majusi and Safavid (although this one could be more contested). The archiver, theoretically, takes an ‘almost’ random sample of Tweets from Twitter (see Wang et al for sampling info).

To determine the location of the Tweeter, I did not use the geodata (as this is rarely used by ppl), but information input by the user themselves on their Twitter account.  Of course there is no way of knowing if this is accurate or not, but for the sake of this analysis, we must assume a significant amount are true.  After filtering out erroneous names, such as people who claim to live in Hogwarts, we were left with around 4500 usable tweets from the original 10,000 tweet sample. I then ran these tweets into Tableau, filtering out duplicate entries (i.e. multiple tweets from the same account). This resulted in about 3,640 unique tweets from unique accounts.


Using Tableau, I first created a cloropeth map that shows the prevalence of sectarian tweets across the region. In the below map, red means a  higher prevalence. As you can see, Saudi Arabia and Egypt appear the reddest, and are the places with the highest number of sectarian tweets in this sample.


Prevalence of Arabic Language Anti Shia Discourse Across Arab World

Yet the above map suggests Egypt and Saudi are alike in terms of tweets. However, the below diagram gives a breakdown of the numbers of tweets, while the surface area of each block represents the proportion of Tweets emanating from each country. As we can see, Saudi Arabia takes the crown with 1,656 Tweets. It is followed by Egypt (420), Kuwait (111), Iraq (71), UAE (56), Yemen (50), Syria (38), Bahrain (36) and Qatar (25) and then Lebanon. (It is worth noting that all countries returned positive hits).


Diagram showing countries with largest number of sectarian tweets

These figures haven’t been controlled for population, or Twitter penetration, the latter of which is difficult to determine (furthermore, figures from the Arab Social Media reports  are perhaps distorted by a large number of bots). Nonetheless, if we are to use these figures from the Arab Social Media Report 2017, we can see that in Egypt that for every 3089 Twitter users, there is one sectarian tweet. In Saudi, there is one sectarian Tweet for 1570 Twitter Users. For Kuwait, as another example, there is one sectarian tweet for every 4504 Twitter users. Thus crudely speaking, the results do not balance out when considering Twitter users, meaning that according to this data, the country with the larger amount of tweets are still the ‘more sectarian’.

More detail

Where possible, I added latitude and longitude points for location input by users. This allowed me to create a map that shows a more detailed breakdown of Twitter users. As you can see from the below map, Arabic, anti-Shia sectarian Tweets are focused on the Arabian Peninsula and Egypt, with the majority occurring in Saudi, specifically Riyadh, Jeddah, and clusters along the Eastern Province. Northern Egypt is also particularly interesting with regards to the amount of discourse.



Anti Shia Sectarian Hate Tweets by Specific Locale (where possible)

Indeed, it would appear from this map, that sectarian discourses online radiate outwards from the middle of Saudi. Of course Riyadh and Jeddah are Saudi’s most populace cities, so it may not be significant in this regard, yet it is interesting to see that Saudi appears to be the centre of this discourse. Also, the rhetoric is almost non-existent across the rest of North Africa and Sudan. It is also not very common in Oman.

The data then, could suggest a number of things:

Online anti-Shia sectarianism are most prevalent in Saudi Arabia and the Gulf.

Such discourses originated in the Gulf and have traveled abroad

Such discourses are much less common outside the Arabian Peninsula, where there are mixed populations with different histories of national struggle

Of course some potential caveats:

Lexically speaking, the terms used for this study may just be a preferred choice used by those Arabic speakers in the peninsula. Sectarianism in other languages other than Arabic would be interesting to explore in a similar way. Similarly, Arabs in others parts of the ME may use different terms, although I am not sure this is the case.

As yet it’s not sure whether long term analysis would yield similar results. Short term, similar sized samples I have done on individual words or phrases leading up to this blog have returned identical results. Inevitably, a longer sample would mostly likely return more and more sectarian tweets from every country outside the Gulf. As these would also likely increase across every country, the proportionality of sectarian tweets would still likely stay the same.

Similar ways of finding other platforms to analyse would be useful, e.g. Instagram and FB. (Although not sure if people would be inclined to be less or more sectarian on different platforms).

To be clear this is not stating sectarianism does not occur everywhere, just maybe that it is more common in some places than others.