Abstract
Terrorist groups have integrated social discourse into their narratives to exploit public fear. The COVID-19 pandemic has provided an opportunity to measure how ISIS modified narratives, stimulated group communication, reinforced enemy hatred, and employed new mobilization/recruitment strategies. Using Significance Quest Theory, this study examines ISIS’s Twitter activity related to COVID-19 (N = 630,201) and how the group leverages the psychological effects and the social disruption. This study employs a two-stage analysis of tweet sentiment and emotion with a statistical analysis of retweet-motivating factors to explore how ISIS manages its narrative, needs, and network. Results show content with higher negative sentiment scores had more retweets across all sample sets. Insights include deeper understanding of extremist activity and group adaptation to external changes.
Introduction
In early 2020, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), known as COVID-19, ignited a global pandemic that continues to linger. Life-altering healthcare protocols to contain the virus has resulted in “stay-at-home” and social distancing policies that restrict public gatherings and travel (OECD, 2020). Social isolation practices, intensified by the fear of infection, have caused many to feel disempowered, frightened, and vulnerable; sentiments which extremist organizations seek to exploit (Kruglanski et al., 2014). As communities vaccinate and develop practices to resume social normalcy, extremists are adapting their strategies to exploit social vulnerabilities by creating messaging that tactically uses the COVID-19 pandemic to promote their narrative among a socially distant and captive online audience (Ackerman & Peterson, 2020).
As post-crisis responses are often less effective at protecting the public, security specialists who rely on knowledge of terrorist psychology and ideology must speculate as to how extremist groups will exploit crises to preventatively counter their messaging. Comerford argues moments of crises serve as “accelerationist” vehicles for extremists, by which they delegitimize the current order and stoke social division (Comerford, 2020). A host of malicious actors have misappropriated COVID-19 as a “wedge issue” targeting minority communities and outsiders, to promote false narratives and conspiracy theories intended to increase support for extremist violence (Comerford, 2020). In this regard, Interpol Secretary General Jürgen Stock emphasized, “Terrorists, like all criminals, have sought to profit from COVID-19, to make money, strengthen their base and to fuel division” (Interpol, 2020).
Since the COVID outbreak, every day in-person social interactions have been limited by restrictions on public gatherings, forcing terrorist operations to adapt to the changing landscape to maintain their relevancy. Groups have adapted messaging focused on public fear and uncertainty while challenging the actions of political institutions (Comerford, 2020). The Institute for Strategic Dialogue’s monitoring of online discourse has found a number of Islamist extremists weaponizing the pandemic with narratives advocating for the establishment of an Islamic state and jihadist violence against non-believers (Comerford, 2020; Institute for Strategic Dialogue, 2020). Veritably, pandemic-era narratives linking and exacerbating existing drivers of radicalization are on the rise, and though this tactic is far from new, there is an emerging interest into how strategies change over time given the pandemic’s duration (Avis, 2020).
As a result, online discourse from platforms like Twitter serve as an insightful resource to assist in our understanding of terrorist operations during the pandemic. ISIS has long recognized the value of online media platforms in addressing various global audiences, as evidenced in Abu Bakr Naji’s “The Administration of Savagery,” posted online in 2004 (Naji, 2004; Wright, 2006). Here, Naji calls for internet-based jihadist “media battles” (2004, p. 73), and ISIS—instructs followers on how to use social media applications and retweets effectively (Al-Rawi & Groshek, 2018). The Shumukh Instigation Workshop (n.d.) video states, “Journalist, you’re a Jihadist,” as the narrator commands, “You need to group all links in a file sharing site; publish the link in the body of the tweet and fill the rest with hashtags”; “Your tweets are your weapons.” Along this strategy, ISIS shares official statements and bulletins from centralized Twitter accounts while publishing live feeds from provincial accounts (Kingsley, 2014). Although social media companies such as Twitter, Facebook, and Google ban extremist content (Andrews & Seetharaman, 2016), allies “are everywhere on Twitter” (Cottee, 2015). At one point, roughly 46,000 Twitter accounts were sympathetic to or supported the group (Berger & Morgan, 2015).
The current study examines some narrative adaptations terrorists employed online during COVID-19, particularly by focusing on non-Arabic keywords on Twitter (“crusader,” “infidel,” and “nonbelievers”), terms often used by radical Islamic militants to convey specific themes (Barker, 2009). Twitter was selected for its popularity, accessibility, impact, and utility by terrorist groups (Choi et al., 2018; Weimann, 2010; Yiğitalp, 2020), with the timeframe from December 2019 to November 2020, a yearlong window on the pandemic. The research utilizes a two-stage analysis: first, examining sentiment embedded in tweets to assess users’ emotions and intent; and second, it uses statistical modeling applied to sentiment, emotions, and other predictors to identify factors associated with retweets.
Online Terrorist Activity During the COVID-19 Pandemic
Twitter is a social networking platform and application that allows users to post and share messages—often embedded with hashtags, URLs, videos, and images—and has been used by extremists to communicate, recruit, and radicalize individuals (Al-Rawi & Groshek, 2018; Berger & Morgan, 2015; Yiğitalp, 2020). Twitter has become a dominant component of ISIS’s social media presence because it provides a medium to disseminate sensationalistic images and videos intended to spread fear and gain support (Yiğitalp, 2020). Recently, ISIS hacked social media links and hashtags, as well as hijacked and repurposed Twitter accounts related to COVID-19, to redirect users to jihadist propaganda (Comerford, 2020; Pauwels, 2020).
With pandemic restrictions keeping more people at home and online, extremists have a larger audience to target their storylines. For example, Sunni jihadists claimed COVID was an anti-Islamic plot, while al-Qa’ida and ISIS contend the virus is Allah’s wrath against nonbelievers and can only be stopped by following “true” Islam (Karadeli, 2020; Shehab, 2020). These propaganda narratives have been deliberately published in English to target vulnerable westerners under stay-at-home orders (Hincks, 2020). The intent of the messaging speaks for itself, yet the fear, uncertainty, and opportunistic claims have attracted users to radical platforms.
Social Engagement During the COVID-19 Pandemic
The rapid spread of the coronavirus ambushed unprepared government institutions and healthcare facilities, triggering overwhelming public dissatisfaction with many government responses (Deane et al., 2021). Many groups exacerbated the growing frustration with conspiracy theories which fueled anti-government extremism, undermined public confidence, and eroded institutional legitimacy (Abuza, 2020; Comerford, 2020; Goodman & Schulkin, 2020; Schaub, 2020; Scheffer, 2020; Wallach & Myers, 2020).
Other extremist organizations have pursued legitimacy-building strategies by engaging in health care projects in poorly governed areas. Groups such as jihadists Lashkar-e-Taiba and Jaish-e-Mohammed in Pakistan (Bellinger & Kattelman, 2020), provided humanitarian services to underscore governmental inadequacies in addressing the COVID crisis. In Afghanistan, the Taliban promoted a public health campaign promising safe passage to healthcare workers treating the virus (Clarke, 2020). In Lebanon, Shi’ite Hizb’allah used the crisis to point out the dysfunction of the Lebanese state while simultaneously crediting their organization for providing medical practitioners and emergency vehicles to assist in pandemic relief (Perry & Bassam, 2020). While these acts may appear benevolent, they are purposely designed to manipulate local support. For extremist groups, pandemic conditions are easily exploited under the veil of altruism.
Increasing Radicalization During the COVID-19 Pandemic
The pandemic’s adverse impact has traversed every corner of the globe, disrupted normal lives while creating feelings of fear and insecurity (Kecmanovic, 2020). A defining element of stay-at-home orders is social restriction, hindering interpersonal social interaction, and increasing social isolation. Together, these conditions carry psychological impacts ranging from anxiety to mild paranoia, and potentially self-destructive behaviors (Cox et al., 2020). Significant loss creates dissociation between individuals’ positive self-perceptions and their painful experiences (McGregor et al., 2001).
Such uncertainties and stressors make people more susceptible to accepting radicalizing narratives that shoulder blame onto “others” and promise neat solutions (McCauley & Moskalenko, 2008). Extremist narratives thus affirm the experiences of those struggling with a diminished sense of self-worth and life meaning (Hogg et al., 2013); while physical distance inhibits the chances radicalization will be detected by others. In a behavioral sense, the living conditions transformed by COVID protocols can be evaluated using the principles of Routine Activities Theory (Felson & Cohen, 1980): motivated offenders have adapted to social restrictions by becoming more active, exposing directed content a larger number of suitable targets, while social isolation reduces the presence of capable guardianship. This links to Significance Quest Theory as a framework for analyzing motives.
Significance Quest Theory for Understanding Extremism
Internet and social media have fueled the rapid global proliferation of violent extremism by introducing a virtual networking dimension for recruitment and radicalization. While researchers assert individual motivations to join such networks hinge on an array of personal and environmental influences, the phenomenon remains studied (Borum, 2011; Doosje et al., 2016; Jensen et al., 2020). From a psychological viewpoint, extremism can be caused by a motivational imbalance wherein one priority becomes central (Gousse-Lessard et al., 2013). Obsessive passion, for a single concern, is distinguished from harmonious passion, for a specific activity balanced with other tasks (Gousse-Lessard et al., 2013; Rip et al., 2012). This view assumes that moderation is the norm while a minority of individuals concentrate their efforts on one interest while suppressing others (Shah et al., 2002).
Significance Quest Theory (SQT), proposed by Kruglanski et al. (2018), employs a psychological model to analyze extremism and is useful for studying its online manifestations. SQT identifies three factors—need, narrative, and network—corresponding to an individual’s desire for significance, the benefits that drive their actions, and their ideological community, respectively. The authors argue the fundamental condition of extremism is a motivational imbalance where a single “need” gains priority, with narratives condoning deadly aggression as a solution. In what follows, we turn our attention to each element.
Need
The need for personal significance is a core factor driving extremism. Kruglanski et al. (2014) posit that all individuals instinctively develop a “need to matter,” often by earning respect and recognition, and this quest for significance resembles a reclaiming of personal meaning, such as healing from negative experiences, loss, or trauma. Meaning-making theories assert individuals act out of frustration to regain a sense of purpose when conditions induce a discrepancy between global and situational meaning (Proulx & Inzlicht, 2012). Though not all downtrodden individuals are predisposed to extremism, this quest for significance increases the likelihood one will seek belonging that provides a sense of purpose and/or an ideology condoning violence as an antidote to feelings of irrelevance.
Young people are particularly susceptible to radicalization and recruitment efforts for reasons such as “adventure” (Bartlett et al., 2010) and risk-seeking (Crenshaw, 1981). To this point, Yiğitalp’s (2020) study of ISIS tweets found messages with violent content had the most followers, while those with softer tones were much less popular. Since tweets receive thousands of views in a short period of time, the communication effect is amplified, especially by retweeting features (Yiğitalp, 2020). Thus, public health restrictions coupled with manipulative narratives likely boost the recruitment potential of extremist organizations.
Narrative
Relatedly, we rely on the importance of meaning-making frameworks to comprehend our own reality. These frameworks shape individuals’ worldviews and values, as the foundations for their identities and life purposes (George & Park, 2016). For political extremists, subscription to an ideological narrative satisfies the need for significance through a commitment to a collective cause, even if the ideology professes the use of violence to garner respect, honor, and appreciation (Kruglanski et al., 2013). Kruglanski et al. (2018) argue significance is uniquely defined by ideological and/or cultural differences embedded in the group’s narrative. In times of relative peace, meaning-making can involve contributions through service, development, accomplishment, and espousing a narrative that defines significance as peaceful support for the cause. During conflict the narrative can advocate for violence against the enemy, conferring glory on martyrs and those who defend the cause.
Network
Lastly, SQT’s network component involves the collective group that subscribes to a narrative. These groups vary from informal assemblies of likeminded individuals to formal organizations devoted to a common purpose (Sageman, 2004, 2008). Networks promoting violence as a means of restoring relevance increases the likelihood individuals will use violence, although individual and situational differences matter. Research accepts people gain personal significance from belonging to social groups they value as important (Tajfel & Turner, 1986), and this need is amplified when alternative needs are suppressed.
In communities where people perceive a threat to their social identity, have experienced social rejection, and/or respond emotionally to these situations, are those likely to promote aggression toward opponents, especially when grievances are perceived as discriminatory, unjust, or humiliating (Smith et al., 2012). Researchers have found this to be true in communities across the globe (Kteily & Bruneau, 2017; Lyons-Padilla et al., 2015) and that it is influential in those marginalized-embracing actions of political extremism (Kamans et al., 2009), to include such violent acts as suicide bombings (Victoroff et al., 2012). Collectively, these studies have identified mechanisms that trigger feelings of insignificance and radicalization, many of which mirror the sentiments behind ISIS’s messaging.
Significance, Emotion, and Terrorism
Using SQT as a framework, this study builds upon previous studies linking the need for significance and terrorism. Terrorist organizations are driven by individuals searching for personal meaning, and ISIS is particularly notorious for these forms of recruitment strategies. Despite being a religiously-focused organization, ISIS managed to attract thousands of foreign fighters (Barrett et al., 2015), many of whom attributed their membership to the desire to escape rejection and alienation and in pursuit for meaning and glory (Callimachi, 2015).
Emotion is a multidimensional human characteristic that reveals individual personality and behavioral traits. People express emotion on various issues triggered by their individual expressive reaction. The American Psychological Association defines “emotion” as a pattern of changes which include physiological arousal, feelings, cognitive processes, and behavioral reactions in response to a personally significant situation (American Psychological Association, 2021). “Happy,” “Sad,” “Anger,” and “Joy” are examples of expressive emotions with sentiment representing a general term of the polarity of emotion (i.e., positive, negative, and neutral). With the growing everyday use of social networks, people express their emotions using social media posts, and Twitter data is a popular choice for text analysis given its limited characters and widespread global use (Sailunaz & Alhajj, 2019). Given these factors, examining emotion and sentiment in individual posts provides us with insight into the reactions of those who post captured messages.
Current Study
While terrorism-linked anxiety has been widely studied (e.g., Carver, 2004; Cheng & Lee, 2010; Mansour, 2018), how extremists exploit global crises remains vague. The effects the pandemic has had on social behavior are meaningful, and given the increased reliance on online communication, the impact of adaptive narratives deployed by groups addressed should be examined. Recognizing the evolving landscape of virtual extremist tactics, it is likely groups will bolster their online approaches to leverage new social interaction dynamics. As such, this study focuses on online extremist activity taking place as COVID spreads-across the globe, exploring the sentiments of tweets and measuring the retweet virality in extremist messaging. The study focuses on three dominant keyword tweets utilizing a SQT framework to evaluate the likelihood messaging would increase extremist attention and support.
Figure 1 presents the study’s conceptual framework in the context of the COVID-19 pandemic (the triggering event) amplifying conditions that radicalize individuals and cultivate terrorist networks. The follower’s hunt for personal significance drives them to propagate extremist messages (“Need”), and ISIS terrorists in particular co-opted the pandemic to spread an accusatory narrative about the virus and the West (“Narrative”). In this study, Twitter is used to capture their narrative, as the platform is used to reach the masses who may or may not have connections to the group. Sheer exposure to the virality of media messaging creates an audience for ISIS’s COVID-19 narrative (“Network”).
Analytical framework.
Based on this, we hypothesize:
H1: Tweets showing negative sentiment will be highly associated with the retweeting of online extreme activity during COVID-19. ISIS narratives with a negative sentiment will be more highly engaged by a network of users.
H2: Tweets conveying negative emotions (i.e., anger, disgust, fear, sadness, surprise) will produce more retweets of online messages during COVID-19. ISIS narratives with a negative sentiment will be more highly engaged by a network of users.
Data and Methods
Data Collection
Twitter is free and public open network in digital communication (Lee & Jang, 2021), making it a useful large-scale data source to understand people’s opinions, social interactions, and mobilization. Terrorist groups, particularly ISIS, utilizes Twitter to disseminate propaganda, justify actions, recruit, threaten, as well as convey support and spread fear (Choi et al., 2018). In this context, social media communications by terrorists can be understood as efforts to facilitate social reaction. Social network features such as followers and retweeting can “validate” sentiments and tensions by spreading hostile beliefs or targeting opponents (Williams & Burnap, 2016).
Twitter data was collected using Twitter’s Streaming Application Programing Interface (API) with a timeframe of a year (December 2019–November 2020). We selected keywords—“coronajihad,” “crusader,” “infidel,” and “nonbelievers”—identified by other researchers as associated with pro-Islamic State coronavirus narratives (Daymon & Criezis, 2020), as well as terms consistently used by ISIS in both hashtags and tweets (Bodine-Baron et al., 2016). For example, on May 28, 2020, ISIS spokesperson Abu Hamza al-Qurashi announced the coronavirus was a punishment on the “crusaders” opposing ISIS. An analysis of the textual narrative, reactions, and amplification of this communication can aid in our understanding of how information is “weaponized” to encourage support.
Keywords originally captured 877,977 tweets, but “coronajihad” was excluded before analysis due to its very limited number (N = 10). While collected tweets were found in different languages, this study focused exclusively on English, thereby retaining 630,056 tweets in our dataset. 1 Within this dataset, data management and pre-processing were performed as Twitter data is inherently textual. We first removed unnecessary words and stop words, symbols (e.g., @), and punctuations. Tokenization was then conducted to parse sentences into a word list. Figure 2 details the data collection and research procedure.
Research procedure.
Analysis Plan
To explore how retweeting of messages was shaped by sentiment and other factors, this study used a two-step integrated analysis plan. Sentiment analysis, one of the most popular opinion mining processing methods in social science, was used first to assess the sentiment and emotion values of tweet content (Mansour, 2018; Zhang et al., 2011). Computational sentiment analysis is commonly used on large-scale data, making it a suitable fit for our study (Balahur et al., 2014). We were particularly interested in two sentiment analysis subsets: (1) eight emotions (anger, anticipation, disgust, fear, joy, sadness, surprise, and trust) and (2) negative and positive sentiments. We utilized a dictionary of sentiments and emotions—contained within the Syuzhet package in R (Jockers, 2015) and the NRC Emotion Lexicon (Mohammad, 2021), respectively—to computationally analyze sentiments, sentiment scores, and categorize tweets by emotion. Each tweet was evaluated by these indicators, and a summed score was assigned to capture each tweet’s overall levels of emotion and sentiment.
The statistical analysis used a series of the negative binomial (NB) models for each group (“crusader,” “infidel,” and “nonbelievers”) and the whole sample was analyzed with Stata version 16. 2 NB regression is a method commonly used to correct for overdispersion caused by risk heterogeneity and event dependence (Nelson, 1980; Winkelmann, 2008). As a result, multicollinearity of the independent variables was not found within the four data subsets. Due to the large size of our data, 3 we ran analyses for both unstandardized and standardized beta coefficients.
Measures
Dependent variable
Our dependent variable is the number of retweets—a continuous, count variable. The number of times an original ISIS post was retweeted provides a dependent measure of ISIS’s narrative flow volume. Consistent with Yang and Counts (2010), who modeled information diffusion in social networks, this volume serves as an indicator of public interest.
Independent variables
We focus on two independent variables, emotions and sentiment, as our initial-stage sentiment analysis. The emotion variables—anger, anticipation, disgust, fear, joy, sadness, surprise, and trust—were measured on a continuous scale while the overall sentiment score is a continuous variable.
Control variables
We controlled for other tweet-functions variables in the models: hashtags, URLs, and videos. Content variables were either coded “1” (i.e., present) or “0” (i.e., absent).
The descriptive statistics are presented in Table 1.
Descriptive Statistics.
Results
Overall Trend
Our overall “keyword tweet” trend is presented in Figure 3. The number of tweets posted by ISIS increased around March 2020, coinciding with the WHO labeling COVID-19 a global pandemic. Perhaps unironically, the three dates with the highest tweet counts in the timeframe are July 11, 2020 (N = 3,190), September 1, 2020 (N = 3,356), and September 2, 2020 (N = 3,056). The first spike occurred 3 days after the July 8th release of a controversial video by a pro-IS group of the U.S. “sinking” in civil unrest and COVID-19. We speculate the spikes on September 1st and 2nd, on the other hand, were triggered by political media debates on the out-of-control civil protests and the slow development of a COVID vaccine. As an observation, many tweets from these dates thematically aligned with the divisive U.S. climate.
Monthly time trend for the number of tweets.
Statistical Analyses
The results of our NB analyses are shown in Table 2A and 2B. 4 As hypothesized, we found negative sentiment motivates retweet activity across both groups, as well as the entire sample. Our baseline model (M1) includes a sentiment measure (sentiment scores only) while M2’s independent variables contain the eight indicators of emotion. In the full sample, every emotion (i.e., anger, anticipation, disgust, fear, sadness, surprise, and trust) except for joy has positive associations with the number of retweets. In each of the samples, however, different emotional traits were found to have stronger predictive effects on the number of retweets. For tweets connected to the word “infidel,” messages conveying anger, fear, sadness, surprise, and (unexpectedly) joy boosted the number of retweets, while anger and surprise showed negative relations to the dependent variable. For “crusader” tweets, anger, disgust, surprise, trust, and anticipation show positive associations with the retweet count, while joy and sadness show a negative association with the dependent variable. For “nonbelievers,” anger, joy, sadness, and surprise show positive associations with the count of retweets, while disgust and fear are negatively associated. While the expression of negative emotions usually predicts a higher number of retweets, positive emotions such as joy and anticipation were likewise predictive.
Negative Binomial Models Predicting Retweet Counts: Crusader and Infidel.
Note. AIC refers to Akaike’s information criterion and BIC refers to Bayesian information criterion.
Models are with unstandardized coefficients.
p < .05. **p < .01. ***p < .001.
Negative Binomial Models Predicting Retweet Counts: Nonbelievers and Full Sample.
Note. AIC refers to Akaike’s information criterion and BIC refers to Bayesian information criterion.
Models are with unstandardized coefficients.
p < .05. **p < .01. ***p < .001.
Building on M2, M3 used sentiment score and emotion measures, and the final model (M4) included control variables. M3’s results were consistent with M2’s in all four samples. M4 consistently affirms the multiplying factor negative sentiment has on the number of retweets, controlling for the presence of URLs, hashtags, and videos. In this model, however, some emotions impacted differently compared to other models. For “crusader” tweets, messages with lower anticipation, fear, joy, and sadness values as well as those with higher anger, disgust, surprise, and trust values have a significant positive relationship to the number of tweets. For “infidel” tweets, the less fearful messages and angrier, sadder, more joyous, and more trustworthy tweets were retweeted more. For the “nonbelievers” sample, tweets rated with lower disgust and fear, and higher anger, joy, sadness, surprise, and trust values were retweeted more.
We discovered interesting patterns in the relationships between sentiment/emotion and retweet counts across the four samples. As hypothesized, high negative sentiment scores yielded more retweets in each sample, supporting H1. Negative emotional expression, in contrast, may have attracted retweets but positive emotions did as well, only partially supporting H2.
Discussion and Conclusions
This study explored the relationships between sentiment, emotion, and the number of ISIS retweets. By analyzing Twitter data collected using three ISIS-linked keywords, as well as a combined sample, our findings underscore the power of negative sentiment to render retweets of extremist content. Eight emotions, classified as negative or positive, showed somewhat different results across the samples, demonstrating sample/keyword distinctions.
Our findings support previous studies on ISIS’s Twitter activity, including the notion that ISIS utilized the pandemic as an opportunity to place blame, indoctrinate, and potentially recruit members (Ackerman & Peterson, 2020). Moreover, negative and positive emotions/sentiments are often hard to separate (Cunliffe & Curini, 2018), accounting for the emergence of mixed findings. In the context of discordant public opinion, our findings suggest external events are inextricably linked to the spread of conflict. Major incidents divide society, with commensurate influence on the success of extremist recruitment. Likewise, COVID-19 produced a context that amplifies the influence of terrorist groups, as ISIS has co-opted with tweets and adherents’ retweets to broadcast messages filled with sentiment and emotion. Critically, messages with stronger retweet power have the greatest potential to reach, and influence, online users.
On a more troubling level, although states established social restrictions to curb the spread of COVID-19, these same measures have yielded a fertile environment for extremist groups to radicalize vulnerable citizens. In addition to feelings of marginalization present prior to the pandemic, fear and insecurity caused by this global crisis exacerbates feelings of uncertainty and psychological symptoms, augmenting susceptibility to radicalism (King & Mullins, 2021; Simon et al., 2021), particularly among youth (Drouin et al., 2020). Furthermore, the increased reliance on online platforms for social contact, engendered by social isolation, weakens the protection of online users by their capable guardians. The combination of these factors provides extremist groups added opportunities to prey on public insecurity and lure more individuals into adopting their beliefs, as many of these organizations have already recognized.
Overall, this research enhances our understanding of online extremist activity in the context of COVID-19, using the emotional and psychological factors of sentiment and emotion as captured from a large-scale Twitter dataset. Across each sample, the results revealed complexities in how sentiment and emotions shape retweet dynamics on Twitter. While this study contributes to the existing fields of security and terrorism studies, criminology, and sociology, it does not exist without limitations that encourage future research. Given the complexity of human emotion, which can be further complicated by the use of sarcasm, detecting correct emotion from text is very challenging and is an area that warrants continued analysis refinement. Sentiment/emotion analysis becomes even more challenging when conflicting emotions are expressed in a single post. The use of dynamic language in such analyses does not lend itself well to the identification of subtle changes, but rather points to the dominant sentiment across the entire dataset. One limitation is our sample was restricted to English-language tweets. Prior research has found that most of ISIS’s messaging is in Arabic; however, it could also be argued the use of English deliberately targets non-Arabic-speaking westerners. In this regard, adding Arabic data could be an approach to expand this study in the future. Second, while a keyword-based tweet collection method drew meaningful data, a broader approach may expand the scope of our understanding. Relatedly, exclusive use of the captured keywords may fail to track the changing nature of these concepts and relationships that take place outside the period of analysis. Finally, future research may benefit from a closer look at online extremist platforms to better understand the nature of prevalent sentiments and emotions.
This study contributes to the growing body of research on extremist groups’ social media tactics during the pandemic era. For policymakers and analysts, our findings provide insight into how narratives are modified and platforms are used to trigger fear and radicalization, whether aimed at current followers or new members. Such information can illustrate communication strategies for mobilization and/or the facilitation of extremist attacks. Lastly, we found sentiments using negative emotions are much more likely to be disseminated to larger audiences, bringing attention for the need of government agencies to proactively educate the public of extremist messaging tactics.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
