The rise of Jihadist propaganda on social networks

Research Article

Abstract

Using a dataset of over 1.9 million messages posted on Twitter by about 25,000 ISIS sympathizers, we explore how ISIS makes use of social media to spread its propaganda and recruit militants from the Arab world and across the globe. By distinguishing between violence-driven, theological, and sectarian content, we trace the connection between online rhetoric and key events on the ground. To the best of our knowledge, ours is one of the first studies to focus on Arabic content, while most literature focuses on English content. Our findings yield new important insights about how social media is used by radical militant groups to target the Arab-speaking world, and reveal important patterns in their propaganda efforts.

Keywords

Computational social science Social media Twitter ISIS Islamic radicalization 

Introduction

Militant groups have long used traditional media and the internet to disseminate information, spread their propaganda [3], and recruit potential militants [4, 7]. But no group to date has been as savvy in terms of its propaganda campaign and recruiting terrorists via the internet, and specifically via social media platforms, as the Islamic State of Iraq and the Levant (ISIS) [27]. ISIS used a panoply of platforms, such as: Facebook, Instagram, Tumblr, Ask.fm, and most prominently, Twitter to spread its message [6]. Previous research has shown that online social media has played an important role in enhancing the popularity of ISIS [13, 14, 29]. The group uses a very successful strategy of having an “online battalion” or the so-called “the mujtahidun (industrious)”, that is likely a small group of 500–2000 active online sympathizers who post and retweet certain tweets to make these messages trending, increasing the group’s exposure and outreach [4]. Previous research worked on understanding ISIS’s social media presence [12, 16, 24, 25, 26], the process of online radicalization that potential recruits go through [5, 10, 31], and how much ISIS gains from such efforts [4, 19].

In this paper, we aim at identifying the topics that ISIS focuses on and investigate how offline events affect ISIS’s online rhetoric. We thus attempt to answer the following two questions: (1) what are ISIS sympathizers talking about on Twitter and how would we categorize their conversations? (2) How are external events reflected in the ISIS Twittersphere? Using a dataset containing all the Twitter posts of about 25,000 ISIS sympathizers, totaling over 1.9 million tweets generated from January 2014 to June 2015, we analyzed the Arabic content of the tweets to answer these questions. The reason for exclusive focus on Arabic content is that the majority of ISIS sympathizers come from the Arab world; Arabic is the most used language among ISIS sympathizers [4, 19]; and prior studies rarely focused on the Arabic content due to the numerous technical challenges. Additionally, Arabic posts constitute over 92% of ISIS activity in our dataset. In this study, we adopt state-of-the-art machine learning tools paired with human annotations to overcome these computational limits. Our study is not the first to be concerned with identifying the topics and rhetoric that radical groups come to focus on and adopt [15, 17, 18, 21].

Smith et al. [28] found that violent groups tend to focus on in-group affiliation, having an impact and confidence in victory. They describe themselves using morality, religion, and aggression rhetoric more than their non-violent counterparts. Our study supports similar hypotheses: we will show that ISIS sympathizers talk mainly about topics related to violence and Islamic theology, with a strong sectarian tone in certain circumstances. Moreover, we will bring evidence that important offline events are strongly intertwined with online conversation, with three particular topics dominating the conversation right before or after ISIS’s activity spikes. For example, we show that increases in the incitement of violence by ISIS supporters and sympathizers on Twitter correlates with certain types of violent offline events; we show a steep rise in the theological talk following ISIS self-proclaiming itself a caliphate; and, we observe an increase in the sectarian tone when the Iranian and Iraqi governments engage with ISIS.

Materials and methods

Data collection

We used a dataset of accounts associated with ISIS sympathizers. The identification of these accounts has been manually performed and validated in two stages: first, a crowd- sourcing initiative called Lucky Troll Club leveraged hundreds of volunteer annotators with expertise in Arabic to identify suspicious accounts that could be tied to ISIS, and reported them to Twitter. The usernames and account IDs of these users were compiled in a publicly available list. Twitter’s anti-abuse team manually verified all suspension requests, and suspended all the accounts related to ISIS based on the violation of Twitter’s Terms of Service policy against terrorist-or-extremist-related activity [13]. Of the Lucky Troll Club we retained only the accounts that have been actually suspended by Twitter via the double-verification mechanism. This yielded a little over 25,000 Twitter accounts associated with ISIS as an object of this study. These users were responsible for posting over 1.9 million tweets from January 2014 to June 2015. All these tweets were obtained through the Observatory on Social Media database (OSoMe) set up by Indiana University [9], which continuously collects the Twitter data stream from the Twitter API (a 10% random sample of the full Twitter data stream). This allowed us to obtain a large sample of the activity of these ISIS accounts, an important technical advantage since studies using small-sample data have shown well known biases [20].

Data analysis

The first step of data cleaning was to retain only tweets in Arabic. This process removed about 8% of the tweets in the original dataset, the remaining 92% being written in Arabic. We then processed the tweets by means of so-called tokenization, using the NLTK tokenizer. Tokenization is the process of splitting a piece of text into individual words, namely word tokens (or tokens in short). The tokens undergo further processing: we used the ISRIStemmer to stem them [30]. Stemming is the process of reducing a word to its root (word stem). This will yield related words to map to the same stem. We removed all non-alphanumeric characters from the stemmed tokens, as well as all English alphabet, both in lower and upper case. In order to do identify the most commonly used stems in the ISIS tweets that have theoretical significance, we removed the stop words (using the stops words from [2]) and constructed a list of the top 100 stems used in the tweets under scrutiny.

Out of the 100 top stems, 66 stems do not convey a clear meaning of what the authors are saying, such as: to go out/exit, land, and do/does, which leaves us with 34 stems that hold significant meanings. We decided to categorize these stems into four categories: Violence, Theological, Sectarian, and Names. The first three are the three main topics we found to be prevalent in the ISIS tweets. The last one, “Names”, includes both names of individuals as well as the Islamic State. In Tables 1 and 2, we provide the list of stems that make up each category, along with their frequency in the whole tweets corpus.
Table 1

Word stems of the violence and theology categories

Violence

Theology

Stem

Translation

Frequency

Stem

Translation

Frequency

قتل

To kill

88,732

خلف

Caliph

80,664

جهد

Jihad

66,268

حسب

Judgment Day

76,116

شهد

Martyr

38,027

حمد

To Thank

70,131

عرك

Fight

36,195

دين

Religion

67,908

حذف

Delete

34,718

كبر

To call for

56,710

حرب

War

29,030

وحد

To unite and it is usually used to express belief in one God

44,225

قصف

Bomb

27,534

شيخ

Shiekh

37,630

فجر

“Fajr” or dawn but also mean to explode

24,149

رحم

To have Mercy

37,369

فتح

Stem for military conquest in a religious sense

22,927

رسل

Messenger

32,847

   

ولي

To rule or to be appointed to rule

30,352

   

شرع

Stem for Sharia

25,916

   

سور

Verse or wall

22,664

Table 2

Word stems of the names and sectarian categories

Names

Sectarian

Stem

Translation

Frequency

Stem

Translation

Frequency

دولة_الإسلامية

Islamic State (ISIS in Arabic/how they liked to be called)

111,892

رفض

To refuse, although here it would be a derogatory term for Shias

66,449

اخبار_الخلافة

The Caliphate News

65,340

صفي

Stem for a derogatory term for Shias

40,939

ولة_الخلافة

The Caliphate State

52,864

كفر

Disbelief or infidel

32,967

دعش

Stem for ISIS

30,336

رتد

Can be the stem for apostasy, also used as derogatory term for adherents of others sects and religions

24,884

اسد

Literally means lion, but here probably referring to Assad, whose name means lion in Arabic

28,201

   

غرب

West

26,475

   

عمر

Omar (name) or age

23,504

   

عرب

Arabs

22,328

   

عاصفة_الحزم

Decisive Storm-military operation led by Saudi Arabic in Yemen

22,153

   

We decided to be as conservative as possible in terms of choosing which stems would be included in these categories. For example, “Violence” only includes stems that incite violent actions. “Theological” includes stems of words that are used frequently in Islamic theology or Fiqh. As for “Sectarian”, we included only derogatory terms of other groups, such as: “Rafdah” (a derogatory term for Shia Muslims) [4] “safiyy” for Safawi (a derogatory term for Shia Muslims and it is the Arabized form of Safavid, the sixteenth century dynasty that established Iran as a Shia State), and “kafara” and “irtad” for infidel and apostasy, respectively, which can be used as derogatory terms for Shias, Yazidis, Christians and even Muslims who do not adhere to ISIS’s vision of Islam. We used this dictionary-based approach to classify each tweet, because we believe that some individuals in our dataset wrote their messages in areas of conflict where the author might have written a short message instead of a longer one due to imminent danger. As such, classifying the tweets rather than the tokens enables us to avoid introducing a systematic bias due to exogenous events.

We built a tweet classifier that counts how many stems are present in each tweet that belongs to the four categories mentioned above. We applied a majority-rule to classify the tweets. For each tweet, the category with the most stems represented in that tweet would be chosen as the category of the tweet itself.1 If a tweet has equal and maximum number of stems from two categories, we classified it as “Other”. Also, if a tweet has no stems from any of the categories, we classified it as “None” and excluded it from future analysis. We isolated tweets classified as “Other,” that contain at least one stem from the “Names” category. Since “Names” is the only category that does not have one defined theme, after removing the tokens classified under this category, we used the majority-rule again to classify the tweets as either “Violence”, “Theological”, or “Sectarian”. We named these categories as “Names + Violence”, “Names + Theological” and “Names + Sectarian” to identify three extra sets of tweets that concurrently identify named entities and one of the three other topics of discussion.

In Table 3, we provide examples of the tweets classified under each category.
Table 3

Examples of tweets labeled under each category

aThe Houthis are a Zaidi Shia-led religious-political movement based in Northern Yemen and is in engaged in ongoing battles with the current Yemeni government and Saudi Arabia

bA gathering congregation hall for Shia commemoration ceremonies to commemorate the martyrdom of Imam Hussein

cTerm used in the medieval ages for Muslims who “strayed from the right path and in believing in monotheism”, could be punishable by death (Lewis 1993)

Results

The classifier we built classified more than half of the tweets (56%) based upon the four categories mentioned in the data analysis section. In Table 4, we report the number of tweets classified under each category, the percentage of tweets classified under each category out of the whole dataset, and the percentage of the tweets classified under each category out of the tweets that were classified in one of the categories. As for the “Other” category, the amount of tweets that we were able to classify under one of the three subcategories “Names + Violence”, “Names + Theological” and “Names + Sectarian”, is 55% (see Table 5).
Table 4

Descriptive statistics for the categorization of the tweets

Categories

Number

Out of the total (%)

Out of the categorized (%)

Names

168,663

8.7

15.5

Other

257,670

13.3

23.7

Sectarian

74,731

3.9

6.9

Theological

387,090

20

35.6

Violence

197,950

10.2

18.2

None

846,912

43.8

 

Total number of tweets = 1,933,016; total number of tweets categorized = 1,086,104

Table 5

The “other” category classified into subcategories

Names + Categories

Number

Out of the total names dataset (%)

Theological

50,098

34.8

Violence

40,829

28.38

Other

33,664

23.4

Sectarian

19,490

13.5

Total number of Names + Categories tweets = 144,081

To represent the volumes of the Tweets and how they change over time, we plotted the ratio of each category to the whole number of the tweets in the dataset (see Fig. 1). Moreover, for the “Names + Category” dataset, we plotted the ratio of each category to the whole number of tweets in this dataset (see Fig. 2). It is important to mention that most of activity happens between January and June 2015, as you can see in Figs. 3 and 4, which show the number of users and tweets by month, respectively.
Fig. 1

Timeline of the ratio of tweets per category over total number of tweets generated by ISIS sympathizers

Fig. 2

Timeline of the ratio of each category in the “Names + Categories” tweets over total number of tweets in this subset

Fig. 3

Number of users per month for each category

Fig. 4

Number of tweets per month for each category

From both figures, it appears evident that the time series fluctuate and spike at different points in time for different thematic categories. Social media literature has reported on correlations between offline and online events in a variety of cases, including during events related to social issues and conflicts [8, 32]. We further investigated to see whether a relationship exists between certain offline events and the Twitter chatter among ISIS sympathizers. We created a list (Table 6) with the most important events related to the four categories of topics we outlined earlier and plotted each category separately with a customized subset of offline events, which are the most relevant to the selected category. Figures 5, 6, 7 and 8 show the ratio of the tweets in the four categories over the total number of tweets by week and the corresponding offline events related to that category. We included similar plots for the “Names + Category” classes in Figs. 9, 10 and 11.
Table 6

Timeline of important external events in both Syria and Iraq

Name

Date

Event description

 

1/5/2014

Fallujah taken by ISIS

 

1/14/2014

Raqqa becoming the capital of ISIS

 

4/1/2015

Iraqi government takes over Takrit

Crucifixions

5/1/2014

Crucifixions in Raqqa

Mosul and Takrit Captured

6/10/2014

Mosul and Takrit taken by ISIS

Iran Deploys

6/12/2014

Iran deploys forces to fight ISIS in Iraq, and helps Iraqi troops regain control of most of Tikrit

Iraq (USA Support)

6/18/2014

Iraq asks the United States to conduct airstrikes against ISIS

Caliphate

6/28/2014

ISIS announces the establishment of a caliphate and rebrands itself as the “Islamic State”

Shaer Gas Battle

7/17/2014

ISIS storms the Shaer gas field and kills 270 people

Sinjar Captured

8/2/2014

ISIS conquers Kurdish towns of Sinjar and Zumar, forcing thousands of Yazidi civilians to flee their homes

 

8/7/2014

President Obama announces the beginning of air strikes against ISIS in Iraq to defend Yazidi citizens stranded in Sinjar

 

9/2/2014

Leaders from ISIS and its jihadist rival, Jabhat al Nusra, meet in Atareb to discuss joining forces. No formal merger or cooperation between the groups is established, but ISIS reportedly sent fighters to help the Nusra Front’s assault on Harakat Hazm, a Western-backed moderate rebel group

Kobani Captured

9/9/2014

ISIS advances on the Syrian border town of Kobani and thousands of refugees flee into Turkey

 

9/22/2014

ISIS spokesman Abu Muhammad al Adnani calls for attacks on citizens of the United States, France and other countries involved in the coalition to destroy the group

 

9/23/2014

The United States launches its first air strikes against ISIS in Syria

 

10/7/2014

The United States significantly ramps up airstrikes in and around Kobani to counter ISIS advances

 

1/7/2015

Two gunmen, Saïd and Chérif Kouachi, attack the offices of French satirical newspaper Charlie Hebdo in Paris, killing 11 people. A third assailant, Amedy Coulibaly, carried out a synchronized attack on a kosher supermarket, taking hostages and killing four people. Coulibaly reportedly declared allegiance to the Islamic State

 

1/26/2015

Kurdish fighters, with the help of US and coalition airstrikes, force out ISIS militants from the Syrian border town of Kobani after a four-month battle

Violent Acts

2/15/2015

ISIS releases a video of Jordanian military pilot Moaz al Kasasbeh being burned alive (2/4/2015)

Libyan militants allied to ISIS release a video showing the beheading of 21 Egyptian Christians, who had been kidnapped on January 12. Egypt launches airstrikes in Libya in retaliation (2/15/2015)

ISIS militants abduct at least 200 Assyrian Christians in northeastern Syria. The US-led coalition launches airstrikes in the same area (2/25/2015)

Takrit Liberated

3/2/2015

The Iraqi government launched a massive military operation to recapture Tikrit with 30,000 Iraqi soldiers and backed by air force

Yazidis Released

4/8/2015

ISIS releases more than 200 captive Yazidis, most of whom had been held captive in northwestern Iraq since mid-2014

Christian Killings

4/19/2015

ISIS posts a video showing militants from its Libyan branch executing dozens of Ethiopian Christians

Palmyra Captured

5/20/2015

ISIS seizes the ancient Syrian city of Palmyra

 

5/21/2015

ISIS militants take full control of Sirte, Libya—Muammar Qaddafi’s hometown

Shite Attack

5/22/2015

ISIS claims responsibility for the suicide attacks on a Shiite mosque in eastern Saudi Arabia, which killed 21 people and injured more than 100

 

6/26/2015

ISIS fighters kill at least 145 civilians in an attack on Kobani, Syria. The same day, ISIS-linked militants attacked a Shiite mosque in Kuwait, killing 27 people and injuring more than 200

We used Glen (2016) as a guide for choosing what events to put in this table

Fig. 5

Tweets classified under “Violence” with important external violent events plotted

Fig. 6

Tweets classified under “Sectarian” with important external sectarian events plotted

Fig. 7

Tweets classified under “Theology” with important external theology-related events plotted

Fig. 8

Tweets classified under “Names” with the announcement of the Caliphate plotted

Fig. 9

Tweets classified under “Names + Violence” with important external violent events plotted

Fig. 10

Tweets classified under “Names + Sectarian” with important external sectarian events plotted

Fig. 11

Tweets classified under “Names + Theology” with important external theology-related events plotted

Robustness check

In order to see the relationship between the volume of the tweets and users and the ratio of the four categories to the volume of the tweets, we plotted eight figures that show each relationship (Figs. 12, 13) using the same temporal resolution (weekly). These plots show that the ratio of Violence and Theological categories fluctuates regardless of the number of users or overall number of tweets in each of these categories, while the rise in Sectarian content does correlate, toward the end, with the rise of users and tweets belonging to the same category. Furthermore, we calculate the standard error (SE) of the mean (from a 100 samples, obtained using bootstrap sampling), showing that the SE is quite small for the tweet and user volumes as well as the categories tweet ratios, reducing the risk of having spurious patterns in our dataset.
Fig. 12

Number of tweets in each category (Violence, Theological, Sectarian, and Names, from top to bottom) per week (blue) vs. ratio of tweets in each category (red in the same ordering as the volume plots). The confidence bands represent the standard error from the mean obtained using bootstrap sampling

Fig. 13

Number of users in each category (Violence, Theological, Sectarian, and Names, from top to bottom) per week (blue) vs. ratio of tweets in each category (red in the same ordering as the volume plots). The confidence bands represent the standard error from the mean obtained using bootstrap sampling

Discussion

More than half of the tweets in our dataset can be classified under the four categories we constructed: this suggests that a significant portion of chatter generated by ISIS sympathizers gravitates around the four issues outlined above. Two topics dominate the conversation, namely theological issues and violence. Looking at the share of these topics in the whole dataset and among those categorized suggests the importance of these two topics for ISIS: theological and violence related issues compose a little over 30% of all the tweets (and more than half of the ones classified). This fact warrants further discussion; thus, we try to draw some possible explanations.

Violence plays a major part in ISIS’s brand and its appeal among ISIS followers. We suggest that ISIS transformed the goal of many Islamist groups. Since the foundation of al-Qaeda, Osama Bin Laden advocated focusing on the “head of the snake”, as he called the USA. He is arguing that once one destroys the source of “evil” that dominates Muslim countries, its puppets (Arab and Muslim leaders) would consequently lose power, and Islamic lands would be freed. This message insisted on gradualism, meaning that he envisioned a battle between Jihadists and the US and its supporters as the battle that would gradually weaken the adversary. On the other hand, ISIS, from the start, refused to portray itself as an underdog, focusing on its victories, atrocious violent acts against minorities (and particularly, Shias, whom they used as a target in order to gain support among certain sectors of the Sunni community), and its call for an Islamic state. This was an intentional posture, which is reflected profoundly in its message to: (1) excite and attract many young men and women to this “exciting” and “victorious” journey that brings pride to its participants, and (2) to escape the spiritual hegemony of al-Qaeda-central and Ayman al-Zawahiri’s attempts to control the jihadi scene in Iraq and Syria.2 Farwell [11] even argues that al-Qaeda was highly displeased with the savage images of killings committed by ISIS and thought it was counterproductive in the long run.

Our results confirm this possible explanation: we see a trend of violence-related rhetoric rising around some of the offline events we outlined. Four events stand out: the crucifixions in Raqqa [1], the invasion of Mosul and Tikrit (Saddem Hussein’s hometown, a place of symbolic importance) by ISIS, the invasion of Sinjar (Yazidi town), and the taking over of the Shaer gas field. All these events, annotated in the timelines, align faithfully with conversation spikes online (see Fig. 5). As for the “Names + Violence” tweets, we see the same pattern of relationship with the crucifixions (see Fig. 9). These results provide support for our prior expectations and theoretical interpretations.

For the theological tweets (Fig. 6), we plotted the announcement of the caliphate and the sectarian events. Here, we expected a sharp rise in theological talk after the caliphate announcement, and suspected that ISIS, after inflicting violence upon minorities, would engage in theological defense and justifications for its actions. In line with what we hypothesized, we see a sharp increase in Twitter discussions among ISIS sympathizers after ISIS self-pronounced itself a caliphate, suggesting how important the perception of this event was among ISIS followers. For the “Names + Theological” tweets, we again see the same sharp rise in tweet volume after the announcement of the caliphate (Fig. 10). It is worth noting that ISIS’s digital magazine Dabiq focused on legitimizing the caliphate since the announcement of its establishment [34]. This should not be surprising, since the concept of the caliphate plays a crucial role in political Islam and arguably in Islamic theology or Fiqh. To many Islamists, Muslims need an “Islamic” state where they can live their lives under the guidance of the Sharia, or in their view free from exogenous corrupting influences. The need for an Islamic state, especially since the collapse of the Ottoman Empire, has been a motivating cause for many Islamic groups, including the groups that some might see as moderate, such as the Muslim Brotherhood.3

For the “Sectarian” category plots (Fig. 7), we can see that the tweets under this category spiked in correlation with events that involves ISIS’s sectarian enemies, most notably the Iranian and the Iraqi governments. For example, around mid-June of 2014, when Iran deployed forces to aid the Shia-dominated Iraqi government take back Tikrit and when the latter asked the US to conduct air strikes against ISIS, we see very visible spikes in the ISIS Twittersphere that include derogatory terms regarding Shias and others in general. Another very clear spike in the usage of such derogatory terms happened the Iraqi government launched a massive campaign to get back Tikrit in March of 2015. For the “Names + Sectarian” class, we see the same pattern: events involving the Iranian or Iraqi government usually coincide with, or slightly precede, spikes in this kind of tweets (Fig. 11). These results support previous research that show that political Islamism and negative views of other religions and groups are highly correlated [33].

Since the “Names” category includes names of various entities, we did not have a uniform expectation regarding the relationship of any specific offline event and the tweets classified under this category, except for the announcement of the caliphate. The “Names” category includes three stems related to the caliphate: a sharp rise in this category after the announcement of the caliphate by the leader of ISIS, Abu Bakr al-Baghdadi, appears as expected (Fig. 8).

Conclusion

The chaos in the aftermath of the US invasion and withdrawal from Iraq and the popular uprising against Assad in Syria in 2011 created a vacuum that enabled groups like ISIS to form. Additionally, there has been evidence in the international relations literature that unstable states are more likely to host terrorist groups than others [22]. But why did ISIS specifically and not another group emerge as the most important and powerful organization in this context? A major reason behind this is its spectacular ability to spread its violent and nihilistic message further and better than any of if its rivals, including established groups like, al-Qaeda. The most prominent way for ISIS to spread its propaganda is through online social platforms, most prominently Twitter. Thus, we attempted, using a dataset of millions of tweets posted by ISIS sympathizers during the one-year timeframe that witnessed ISIS’s rise, to capture what these sympathizers are talking about, what message they wanted to convey and how events on the ground affect the Twittersphere.

We concluded that violence, Islamic theology, and sectarianism play a crucial role in ISIS messaging. In some cases, ISIS emphasis on some topics of discussion slightly anticipated events on the ground: for example, the use of sectarian language online toward those entities perceived as adversaries can be observed prior to executions and attacks. In other cases, ISIS focused on certain topics as an aftermath of offline events: this was the case, for example, when ISIS inflicted violence upon minorities, and then engaged online in theological defense and justifications for its actions. Possibly, the most prominent event during this period, in terms of its importance and perceived meaning to ISIS sympathizers, was the announcement of the caliphate. This event was both preceded and followed by several shocks in the Twittersphere, with multiple spikes occurring across different categories of discussion shortly before and slightly after the event. In conclusion, our work sheds light on the ability of ISIS to systematically and programmatically corroborate its agenda with remarkable coordinated activity on social media.

Footnotes

  1. 1.

    The percentages of tweets that were labeled under each of the categories and had no stem words from the other three categories ranged from 90 to 96%. This means that most of the tweets that were labeled at this step had stem words just from one category and zero from the others and that majority-rule works well in this situation.

  2. 2.

    For more on the subject, read chapter eight “The AQ-ISIS War” in [4].

  3. 3.

    For more on the subject, read Milestones [23].

References

  1. 1.
    Abdelaziz, S. (2014). Death and desecration in Syria: Jihadist group ‘crucifies’ bodies to send message. CNN, 2 May 2014.Google Scholar
  2. 2.
    Alajmi, A., Saad, E., & Darwish, R. (2012). Toward an arabic stop-words list generation. International Journal of Computer Applications, 46(8), 8–13.Google Scholar
  3. 3.
    Allendorfer, W. H. & Herring, S. C. (2015). ISIS vs. the US government: A war of online video propaganda. First Monday, 20(12).Google Scholar
  4. 4.
    Berger, J. M., & Morgan, J. (2015). The ISIS Twitter census: Defining and describing the population of ISIS supporters on Twitter. The Brookings Project on US Relations with the Islamic World, 3(20), 1–65.Google Scholar
  5. 5.
    Bermingham, A., Conway, M., McInerney, L., O’Hare, N., & Smeaton, A. F. (2009). Combining social network analysis and sentiment analysis to explore the potential for online radicalisation. In International conference on advances in social network analysis and mining, 2009. ASONAM’09 (pp. 231–236). IEEE.Google Scholar
  6. 6.
    Bodine-Baron, E., Helmus, T. C., Magnuson, M., & Winkelman, Z. (2016). Examining ISIS support and opposition networks on twitter. Technical report, RAND Corporation Santa Monica United States.Google Scholar
  7. 7.
    Cohen-Almagor, R. (2013). In internet’s way: Radical, terrorist Islamists on the free highway. International Journal of Cyber Warfare and Terrorism, 2(3), 39–58.CrossRefGoogle Scholar
  8. 8.
    Conover, M. D., Ferrara, E., Menczer, F., & Flammini, A. (2013). The digital evolution of occupy wall street. PLoS ONE, 8(5), e64679.CrossRefGoogle Scholar
  9. 9.
    Davis, C. A., Ciampaglia, G. L., Aiello, L. M., Chung, K., Conover, M. D., Ferrara, E., et al. (2016). OSoMe: The IUNI observatory on social media. PeerJ Computer Science, 2, e87.CrossRefGoogle Scholar
  10. 10.
    Edwards, C., & Gribbon, L. (2013). Pathways to violent extremism in the digital era. The RUSI Journal, 158(5), 40–47.CrossRefGoogle Scholar
  11. 11.
    Farwell, J. P. (2014). The media strategy of ISIS. Survival, 56(6), 49–55.CrossRefGoogle Scholar
  12. 12.
    Ferrara, E. (2017). Contagion dynamics of extremist propaganda in social networks. Information Sciences, 418–419, 1–12.CrossRefGoogle Scholar
  13. 13.
    Ferrara, E., Wang, W.-Q., Varol, O., Flammini, A., & Galstyan, A. (2016). Predicting online extremism, content adopters, and interaction reciprocity. In International conference on social informatics (pp. 22–39). Berlin: Springer.Google Scholar
  14. 14.
    Fisher, A. (2015). How jihadist networks maintain a persistent online presence. Perspectives on Terrorism, 9(3), 3–20.Google Scholar
  15. 15.
    Halverson, J. R., Goodall, Jr., H., & Corman, S. R. (2011). In Master narratives of Islamist extremism (pp. 137–148). Berlin: Springer.Google Scholar
  16. 16.
    Klausen, J. (2015). Tweeting the Jihad: Social media networks of western foreign fighters in Syria and Iraq. Studies in Conflict & Terrorism, 38(1), 1–22.CrossRefGoogle Scholar
  17. 17.
    Lee, E., & Leets, L. (2002). Persuasive storytelling by hate groups online: Examining its effects on adolescents. American Behavioral Scientist, 45(6), 927–957.CrossRefGoogle Scholar
  18. 18.
    Leuprecht, C., Hataley, T., Moskalenko, S., & McCauley, C. (2010). Winning the battle but losing the war? Narrative and counter-narratives strategy. Perspectives on Terrorism, 3(2), 42–57.Google Scholar
  19. 19.
    Magdy, W., Darwish, K., & Weber, I. (2016). #FailedRevolutions: Using Twitter to study the antecedents of ISIS support. First Monday, 21(2). ISSN 13960466. http://firstmonday.org/ojs/index.php/fm/article/view/6372.
  20. 20.
    Morstatter, F., Pfeffer, J., Liu, H., & Carley, K. M. (2013). Is the sample good enough? Comparing data from twitter’s streaming API with Twitter’s Firehose. In International Conference on Weblogs and Social Media (ICWSM) (pp. 400–408).Google Scholar
  21. 21.
    Payne, K. (2009). Winning the battle of ideas: Propaganda, ideology, and terror. Studies in Conflict & Terrorism, 32(2), 109–128.CrossRefGoogle Scholar
  22. 22.
    Piazza, J. A. (2008). Incubators of terror: Do failed and failing states promote transnational terrorism? International Studies Quarterly, 52(3), 469–488.CrossRefGoogle Scholar
  23. 23.
    Qutb, S. (1964). Ma’alim fi-l-tariq (milestones). Cairo: Dar al-Shuruq.Google Scholar
  24. 24.
    Rowe, M., & Saif, H. (2016). Mining pro-ISIS radicalisation signals from social media users. In International Conference on Weblogs and Social Media (ICWSM) (pp. 329–338).Google Scholar
  25. 25.
    Saad, S., Chamoun, M., & Bazan, S. B. (2015). Infowar on the web: When the caliphate goes online. In Proceedings of the ACM web science conference (Art ID. 53). ACM.  https://doi.org/10.1145/2786451.2786487.
  26. 26.
    Saif, H., Dickinson, T., Kastler, L., Fernandez, M., & Alani, H. (2017). A semantic graph-based approach for radicalisation detection on social media. In European semantic web conference (pp. 571–587). Berlin: Springer.Google Scholar
  27. 27.
    Shane, S., & Hubbard, B. (2014). ISIS displaying a deft command of varied media. The New York Times, 30, 30 Aug 2014.Google Scholar
  28. 28.
    Smith, A. G., Suedfeld, P., Conway, L. G., III, & Winter, D. G. (2008). The language of violence: Distinguishing terrorist from nonterrorist groups by thematic content analysis. Dynamics of Asymmetric Conflict, 1(2), 142–163.CrossRefGoogle Scholar
  29. 29.
    Stern, J., & Berger, J. M. (2015). ISIS: The state of terror. New York: HarperCollins.Google Scholar
  30. 30.
    Taghva, K., Elkhoury, R., & Coombs, J. (2005). Arabic stemming without a root dictionary. In International conference on information technology: Coding and computing, 2005. ITCC 2005 (Vol. 1, pp. 152–157). IEEE.Google Scholar
  31. 31.
    Torok, R. (2013). Developing an explanatory model for the process of online radicalisation and terrorism. Security Informatics, 2(1), 6.CrossRefGoogle Scholar
  32. 32.
    Varol, O., Ferrara, E., Ogan, C. L., Menczer, F., & Flammini, A. (2014). Evolution of online user behavior during a social upheaval. In Proceedings of the 2014 ACM conference on Web science (pp. 81–90). New York: ACM.Google Scholar
  33. 33.
    Weber, I., Garimella, V. R. K., & Batayneh, A. (2013). Secular vs. Islamist polarization in Egypt on Twitter. In Proceedings of the 2013 IEEE/ACM international conference on advances in social networks analysis and mining (pp. 290–297). New York: ACM.Google Scholar
  34. 34.
    Winter, C. (2015). Documenting the virtual ‘caliphate’. Quilliam Foundation, 33, 1–50.Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  1. 1.Political Science and International Relations DepartmentUniversity of Southern CaliforniaLos AngelesUSA
  2. 2.Information Sciences Institute, University of Southern CaliforniaMarina Del ReyUSA

Personalised recommendations