This is the online version of the article. To access a print version with page numbers for citation and reference purposes, select "Download" to the right and then choose "Formatted PDF."
Online discussion forums have been identified as an online social milieu that may facilitate the radicalization process, or the development of violent narratives for a minority of participants, notably youth. Yet, very little is known on the nature of the conversations youth have online, the emotions they convey, and whether or how the sentiments expressed in online narratives may change over time. Using Life Course Theory (LCT) and General Strain Theory (GST) as theoretical guidance, this article seeks to address the development of negative emotions in an online context, specifically whether certain turning points (such as entry into adulthood) are associated with a change in the nature of sentiments expressed online. A mixed methods approach is used, where the content of posts from a sample of 96 individuals participating in three online discussion forums focused on Islamic issues is analyzed quantitatively and qualitatively to assess the nature and evolution of negative emotions. The results show that 1) minors have a wider range of sentiments than adults, 2) adults are more negative overall when compared to minors, and 3) both groups tended to become more negative over time. However, the most negative users of the sample did not show as much change as the others, remaining consistent in their narratives from the beginning to the end of the study period.
In an exclusive interview with CNN on December 21, 2015, Abu Hurriya – a former self-proclaimed "chief propagandist for Al Qaeda in the United States" (Cohen & Goldschmidt, 2015, para.11)—admitted “he was once a lost, angry young man (a "seeker") who went through the radicalization process, just as young people are going through it today”. According to Harriya, who was speaking to the effects of the Internet as a recruitment tool, “I can understand how [youth] can get to that point. They're young and vulnerable” (Cohen & Goldschmidt, 2015, para.12). Before entering into the radicalization phase, Hurriya states he was lost and angry. It is this emotional state of anger and frustration (as well as the strain from feeling lost) that appear to be a critical aspect of a youth’s decision to seek out other like-minded individuals who could make them feel connected.
The role of the Internet in the process of radicalization is the object of an increasing amount of scholarship (for a recent review, see Ducol, Bouchard, Davies, Ouellet, & Neudecker, 2015; Neumann, 2013; Ryan, 2007). Many of the most informative studies use cases of known terrorists and look back through their trajectories, the way Hurriya did above, to find whether or how the Internet played a role in the process. The inherent sample selection bias of those studies (i.e., only confirmed terrorists are selected) is usually compensated by the benefits of studying a subpopulation of individuals who moved from radical ideas to actual violent acts. A complementary, but rarely investigated aspect of this issue is to look at a much broader set of individuals and their online involvement, even before any traces of radicalization are present. In other words, we still lack a detailed understanding of the nature of the conversations that young individuals have online, and the emotions they seem to convey. Negative emotions, or manifestations of frustration, have been associated with the development of violent and extreme ideas (Fredrickson, 2004), but have also been an integral predictor of violent criminal behavior and delinquency more generally, via General Strain Theory (Agnew, 1992). Thus, it becomes necessary to understand whether frustration and anger can be captured in the online setting both qualitatively and quantitatively and whether it is possible to assess its evolution over time.
Drawing from Life Course Theory (LCT) and General Strain Theory (GST), the current study will seek to address the development of negative emotions in the online context, specifically, whether certain turning points, such as entry into adulthood, are associated with a change in sentiment expressed online. The case study selected for analysis is a series of online discussion forums frequented by English speaking Muslim youth. This group of youth was chosen as they have faced, more so than other groups, a tumultuous period where a specific interpretation of their religion has been associated with terrorist activities. The vast majority of youth analyzed were not reporting any activities that could be associated with radicalization. Instead, the presence of frustration, as expressed in these youths’ narratives online, was examined with a focus on changes associated with the transition to adulthood. Thus, the current study is not a study of radicalization but can inform such literature indirectly—through an examination of the range of sentiments expressed by youth online.
On average, youth between the ages of eleven and eighteen tend to devote about eleven hours a day exposing themselves to electronic media and overall social networking sites (Shapiro and Margolin, 2014). The psychosocial impact this amount of time spent using electronic forms of media is having on adolescents today has yet to be fully understood, especially in relation to individual maturation process (Ducol et al., 2015; Shapiro and Margolin, 2014).
The general topic of youth development in the more traditional offline context has been thoroughly researched by scholars and practitioners in various fields including, but not limited to, psychology (Cassell, Huffaker, Tversky, and Ferriman, 2006), criminology (Loeber, 1990; Donker, Smeenk, and van der Laan, 2003), psychiatry (Guan and Subrahmanyam, 2009), medicine (Shek and Yu, 2011), and social sciences more generally (Berson and Berson, 2005). Regardless of the discipline, a common theme that emerges seems to be the importance, and significance, of transitioning into adulthood (or developing throughout the general time period of late teens/early twenties) on youth behaviour (Cernkovich and Giordano, 2001).
The emergence of the Internet has brought with it new complications and facets for youth development, with some academics suggesting this new phenomenon to be detrimental to the maturation process (Guan and Subrahmanyam, 2009), and others promoting its ability to “enrich and extend life experiences” (Berson and Berson, 2005, p.29); but most believing the Internet to serve both positive and negative roles (Subrahmanyam and Greenfield, 2008). One facet that most academics will agree upon is that of youth vulnerability in the online setting (Ryan, 2007; Berson and Berson, 2005; Young, 1998; McKenna and Bargh, 2000). According to Berson and Berson (2005), “the permeation of the Internet into the lives of youth can expose them to information with questionable legitimacy, ideas that can be contrary to positive behaviors, and messages that are intended to manipulate their actions or beliefs” (p. 30).
While the risk factors for youth engaging in online interactions and activities have been relatively well researched and documented (Ryan, 2007; Berson and Berson, 2005; Young, 1998; McKenna and Bargh, 2000), the developmental aspects of youth maturation (transitioning from adolescence into adulthood) have yet to be comprehensively examined in relation to the online setting. Additionally, some discuss the effects of Internet exposure on youth development offline – specifically that of online pornography negatively impacting adolescent sexual development (Ybarra and Mitchell, 2005), as well as online risk factors more generally (Guan and Subrahmanyam, 2009)—but do not address the extent to which youth development can be manifested online. Thus, the current study will seek to fill this gap in the literature by drawing focus to individual development over time, online.
To understand how frustration and anger can potentially be related to violent behavior, we turn to Robert Agnew's General Strain Theory (GST) for guidance. Anger is a core emotion for GST and is considered by Agnew (1992) to be a key factor in the development of criminal behavior (Froggio, 2007). Anger is said to be a result of experiencing subjective strain whereby negative emotions accumulate, and feelings of intense negativity are produced. It is important to note, however, that the extent to which a negative emotion is experienced (if at all), is highly dependent on the individual and their attributes. For some, an emotion such as anger has the ability to "increase the individual's level of perceived injury, create a desire for revenge, motivate action and lower inhibitions" (Froggio, 2007, p.389), and has been said to result from underlying feelings of intense frustration and aggression (Berkowitz, 1989; Dollard, Doob, Miller, Mowrer, & Sears, 1939). Thus, it becomes important to understand the extent to which extreme negativity is taking place in online forums, as this negativity may reflect the development of frustration, anger, and even violence.
Our study specifically examines the transition of minors entering into adulthood, which can be particularly stressful for many youth, both in their academic and personal lives. The ability to take on a new role as an adult, make your own choices and live your life however you please (without the direct interference of parents/authority figures), brings with it the inevitable experiencing of new stressors (Agnew,1997). Agnew (1997) makes note of the vulnerability accompanying the years of adolescence, since the cognitive capabilities are in the process of developing, but have yet to fully form – meaning, youth have a limited ability to see the ‘bigger picture’ and fully comprehend the situations they find themselves in. This inability to manage (or form effective coping strategies) and identify the source of stressful situations is considered by Agnew (1997) to be an additional strain on the youth, and allows the individual to experience negative emotions – the least of which are frustration and anger (Froggio, 2007, p.411).
Over the years, the manner in which the radicalization process is understood has changed and developed to include both offline and online factors. Sampson and Laub’s Life Course Theory (LCT), emphasizes the importance of two fundamental concepts: trajectories and transitions. Trajectories are synonymous with individual pathways throughout the lifespan, developing and changing based on various transitions and turning points. Transitions, as defined by Elder (1985), "are marked by specific life events that are embedded in trajectories and evolve over shorter time spans" (Sampson and Laub, 1993, p.254). These transitions are predominantly age-graded—life events such as entrance into high school or university, marriage, parenthood, sickness, and or death; although, age may not always be a factor as the timing of one's life events may be early or late.
The current study seeks to better understand the emergence and development of extreme negativity in the online setting. Part of that task requires a deeper understanding of the various pathways and extraneous factors that may influence an individual’s sentiment trajectory. The impact certain life events, or transitions, can have on a life trajectory is supported by Sampson and Laub’s (1990) study on transitions into adulthood, and the social ties associated with it – specifically in relation to university education, marriage, and job stability.
Sampson and Laub (1990) stipulate that social bonds developed in adulthood seem to have a significant effect on modifying (or reducing) criminal and deviant behaviour during adolescence. While youth tend to age out of certain life behaviors and mentalities, the same cannot necessarily be said for that of negativity (or negative affect). Literature to date is mixed on this topic, with some studies suggesting negative affect to increase with age (Ferring and Filipp, 1995), others finding support for a decrease in negative affect over time (Mroczek and Kolarz, 1998 ;), and several reporting no changes at all (Smith and Baltes, 1993).
The current study will employ several concepts drawn from LCT and GST to address the development of negative emotions in the online context – specifically whether entry into adulthood is associated with a change in the nature of sentiments expressed online. The research question will be answered via a mixed methods approach. The extent to which individuals vary in their sentiment over time will be determined quantitatively via the use of a sentiment analysis software that attributes specific scores to narratives from the extremely positive to extremely negative (Kennedy, 2012). After graphing each user’s sentiment trajectory, those who display significant peaks or drops in their trends will be extracted for further analysis, in addition to those displaying extreme negativity of a consistent nature. The content of user posts during these peaks and/or drops in sentiment will be analyzed qualitatively to extract the themes emerging from their online narrative over time – themes related to turning points and life transitions, most notably that of entrance into adulthood. Finally, the connection between the quantitative and qualitative analysis of user posts will be examined.
Since the events of September 11 2001, young Muslims have learned to develop their identity in a society that typically associates their religion with major terrorist events around the world. For this reason, three forums both open to the public and focused on Islam-related issues were selected: Islamic Awakening (IA), Shia Chat, and Ummah. Although the study aims are broader, known terrorists have also named Islamic Awakening as a forum they frequent (Berger, 2011). The research was purely non-participant observation: a public forum was used (no password or login needed), and we did not intervene in any of the conversations. We also anonymized user pseudonyms for this paper.
Data was collected using a software MITS Crawler, which is a custom-created web-crawler designed in part to capture the content posted to openly accessible discussion forums. The software captures information from a user-selected forum by downloading all its webpages, parsing the page apart with the use of forum-specific ‘rules' to capture all the useful information present on the forum, then storing it in a corresponding database. The database is designed to resemble the structure inherent to discussion forums and is navigated in the same way: each forum has many sub-forums, which in turn have many threads of discussion, each with at least one post. The result is a collection of copied forums that, if possible, are updated daily with the new posts from the original discussion forum.
Given limited time and resources, we introduced several restrictions to select the sample for analyses. First, we restricted our analyses to forum users associated with a single country (Canada), allowing us to partially control for the variety of socio-political contexts experienced by the international crowd encountered on the forums. Second, we divided our sample in two, to compare a subsample of users who started on the forum during their adolescence with others who started as adults.
Users were selected from all three forums (Islamic Awakening, Shia Chat, and Ummah) in varying ways, as each forum presented unique restrictions surrounding the ability to search for members, determine their age, or establish current/previous geographic location. Shia Chat was the most liberal with their privacy settings, allowing for members to be identified based on geographic location, or even simple keyword searches (such as “Canada”, “Canadian”, “Cana”, “West”, etc.). In addition, Shia Chat had a sub-forum entitled “Introduce Yourself Here!” whereby thousands of members introduced themselves, commonly revealing detailed sociodemographic information – including real name, age, current geographic location, birth location, relationship status, and other personal information. Manual analysis (i.e., reading through the threads and individual comments) allowed for over 250 users to be identified and extracted for consideration in the initial stages of the study.
The Islamic Awakening and Ummah websites were slightly more challenging to navigate, as both forums restricted their privacy settings to the extent of being unable to search for/view members based on keywords or variables. For these forums, the MITS Crawler was employed to search for users based on specific keywords (identifying their Canadian tie and approximate age). Based on the information extracted by the crawler, it became feasible to estimate the ages of the individuals in question by examining the content of the posts being made. For example, some individuals overtly stated their ages, making it possible to deduce the age while taking into consideration the date (year) in which the post was made. Others posted information pertaining to their grade or year in high school or university, or birthdays they had recently celebrated. When the age was not specifically stated by the user, it became necessary to corroborate information (like an individual being in the first year of university) with additional information found on the same individual (like another user wishing them a "happy birthday", or other statements made concerning an age bracket).
Of all users identified and extracted in the preliminary stages of the study, a total of 96 individuals who either self-identified themselves as Canadians or had mentioned living in Canada at one point in their lives, were sampled from these forums. Forty-eight (50%) of these individuals started on their forum before the age of 19—they were considered the sample of "minors." Forty-eight Canadian minors and forty-eight Canadian adults (as a control group) were sampled from the forums IslamicAwakening, Shia Chat, and Ummah. All users sampled were required to meet the following criteria: 1) be on the forum for at least 12 months 2) have at least 100 posts. The 96 users in the sample were the only ones fitting these criteria. Additionally, those who were considered minors were required to meet the following criteria: 1) they needed to be under the age of 18 at first post, and 2) active on the forum through the transition into adulthood; whereas those who were considered adults needed to be over the age of 18 at first post.
The variables “years on forum”, “number of posts”, “age at join”, and “age today” were extracted for individual users (based on profile details made available by each forum), and calculated so as to compare and contrast between the minors and adults groups. “Age at join” was extracted either directly from the forum (if available), or searched for using a keyword search via the MITS Crawler. “Age today” was simply a calculation of “age at join” plus the appropriate years to reach 2015 (time of capturing data).
Posts for each user sampled were extracted through the MITS Crawler. The posts were then used to conduct sentiment analysis and average sentiment scores per 6-month time periods of analysis. In total 282,411 posts were extracted and analyzed. The time period whereby posts and users were extracted from each forum was between 2002 and 2015 inclusive (14 years). The year 2002 was chosen as a start date since this was the year Shia Chat opened as a public forum and was the earliest date from which users and posts could be extracted.
For both minors and adults, the majority of users originated from ShiaChat, followed by Ummah, and IA. For adults, however, Ummah forum members seemed to be more active, on average, than those found on the other two forums. Approximately 68.76% (or 33/48) minors continued to post after they became an adult.
A sample of posts was subjected to qualitative analysis. The online MITS crawler, alongside manual searching, was used to search for keywords and phrases in individual user trajectories to determine transitions the user already entered into (such as adulthood or university), and transitions they were currently entering into. Transitioning into adulthood was defined as turning 19 years of age, and entering into university was determined by the user explicitly stating their involvement in, and the start of, university. In doing so, keywords such as "university", "school", "age", "birthday" (and many others), were searches for with the crawler. Qualitative coding (and analyzing) of individual posts occurred for each user, within 6-month intervals – to match that of the quantitative, sentiment analysis. A random sample of posts during this time period was analyzed, making sure to identify any turning points and transitions.
Sentiment score is a measure that quantifies the emotionality and positivity/negativity of content. This measure provides a quantitative understanding of the content of information being found in online forums – specifically, the extent to which positive and negative sentiment is present. The sentiment analysis program producing the score aims to automatically extract the emotions or attitude of a text, or narrative, and assign a value that ranges from the ‘negative’ to the ‘positive’ (Kennedy, 2012). For example, a word such as “love” might incur a positive sentiment score of 20 (approximation), whereas a word such as “hate” might incur a sentiment score of -20 (approximation).
SentiStrength is a computer program that analyzes scripts of sentiment (verbal discourse) and attributes a score to each word based on a polarized scale of positive to negative connotations associated with each word (Kennedy, 2012). This program has a built-in dictionary whereby every word has a pre-determined score of positivity or negativity, and as such, will average out the sentiment score for the portion of data being examined. Text analysis is incredibly useful at both the micro and macro level as it allows for “thousands of posts [to] be sifted [through] and wider trends and dynamics [to] be discovered” (Neumann, 2013, p.450).
Individual sentiment scores were determined for each post collected over the years spent on the forum, averaged into one score per 6-month time periods of analysis (first 6 months in the year, and second 6 months in the year), and then sorted by date in order to plot these points on a chart and examine the trend in sentiment over time.
The current study uses both quantitative and qualitative approaches to track the evolution of sentiments over time. The analysis proceeded in two main steps: 1) Before/After analysis of sentiments using SentiStrength; and 2) Analysis of sentiment consistency by individual, drawing from qualitative analysis of individual narratives. A final, secondary step is to compare the individuals deemed most negative from the quantitative analysis of sentiments to those deemed to have the most violent narratives from the qualitative analysis.
The goal of before/after analysis is to better understand the extent to which user sentiment trends develop over time, specifically whether or not the transition into adulthood encourages a change in sentiment (whether that be an increase in positivity, an increase in negativity, or neither). For the users who started as minors, the before/after analysis will test whether there is a statistically significant difference between a minors sentiment scores during their youth (before-19 years old) compared to their sentiment scores after they turn into an adult (after-19).
For comparison purposes, the adults group will be examined as well. Since this group does not have the turning point of adulthood (as they began on their forum as an adult), their first 50% of posts will be compared to their second 50%. A Paired Sample T-test will be conducted on 33 minors who had both before and after sentiment scores, as well as the 48 adults with their first/last 50 percent of sentiment scores.
Sentiment ranges were generated for each minor (n=48) and adult (n=44), and then the median determined from the list of ranges (minors median=4.95; adults median=8.32). By using the median (and not the mean, which was overly impacted by outliers), it was possible to differentiate between consistent and inconsistent trends in this sample. The adults group had four individuals who did not possess enough data points to generate a range; thus, the total number of adults used for the current analysis was 44.
To determine the sentiment trajectory, the median range of sentiment per individual was used. If an individual's range fell outside this standard number (4.95 for minors, 8.32 for adults), then it could be understood that their sentiment scores were varying to a degree that would classify their trajectories as being "inconsistent". If an individual's range fell within this standard number, then it could be inferred that the individual had a somewhat "consistent" trajectory.
The qualitative component of the current study employed directed content analysis as a coding technique. This qualitative coding scheme was chosen for its ability to best interpret and reveal meaningful information from the data examined. Directed content analysis begins with a theory or findings established from prior research, allowing the coding to be guided as per the pre-established direction of the theory/research (Hsieh and Shannon, 2005).
While LCT traditionally measures rates and trajectories of crime and delinquency over the life course, the current study substitutes these variables for that of negativity and positivity in the content of their online posts. We approached this analysis with the assumption that certain transitions and turning points may be associated with changes in the levels of negative discourse online. We sought to analyze users’ narrative to investigate the presence of frustration and negative emotions. In doing so, we paid particular attention to the consistency and inconsistency displayed by individual users, were we able to perceive distinct changes in tone and content over time.
All users who displayed sentiment trajectories that were considered to be above the sample average for negativity were considered for further qualitative analysis, after which content analysis was conducted to identify turning points evident in each user's online trajectory, as well as their effect on individual sentiment trends. Only those displaying negativity in their sentiment trajectory were considered since the current study sought to address the development of negative emotions in the online context, as opposed to positive emotions – a limitation for the study. Furthermore, qualitative analysis allowed for former classifications (based on sentiment trends) to be assessed.
The online MITS crawler was used to search for keywords and phrases in individual user trajectories. Based on the posts observed throughout the year under examination, it was possible to determine transitions the user already entered into (such as adulthood or university), and transitions they were currently entering into, through manually searching for keywords/events within all of their posts. Analyzing these turning points was conducted through manually reading user posts over the course of their time on each forum to identify appropriate timing and determine levels of negative speech in relation to said turning point and transition. In some instances, working retrospectively based on statements made in future posts was required. The following two turning points/transitions were determined and considered relevant for examination: entrance into adulthood and entrance into university.
Qualitative coding (and analyzing) of individual posts occurred for each user, within 6-month intervals. A random sample of posts during this time period was analyzed, making sure to identify any turning points and transitions. Furthermore, additional important factors such as the development of interests/opinions and beliefs, the number of posts exhibiting frustration or anger throughout time on the forum, and any other key information (posts) that may have been of interest and relevant to the study were recorded. Once all information was collected, it was possible to qualitatively understand the development of user sentiment over time by contextualizing statements (or posts) made by each user. In some instances, it was necessary to retrospectively read through the thread to gain a better understanding of the post made by the user under analysis – in other words, understanding the true nature of the post based on context (a factor that is unable to be accurately accounted for by SentiStrength).
To begin, the sample of minors and adults are compared. An independent samples T-test was conducted to determine whether or not statistically significant differences could be found between the two groups (Table 1).
Table 1. Descriptive statistics
Years on forum
Number of posts
Age at join
Years on forum
Number of posts
Age at join
Minors and adults
Years on forum
Number of posts
Age at join*
Note: Welch`s t-test used (equality of variance not assumed), comparing minors and adults; *p<.01.
First, the variables “years on forum” (14 years on average) and “number of posts” present similar numbers for both minors and adults, suggesting consistency between the two groups. These similarities imply that potential differences in sentiments between the two groups would not necessarily be explained by these variables. Second, as expected given the research design, age at which they joined the forum, and their age today were significantly different.
Each user was attributed an overall sentiment score based on the average of all sentiment scores calculated by post. Descriptive statistics were run on these overall sentiment scores for both minors and adults – see Table 2. As can be seen, adults appear to be significantly more negative overall, when compared to minors, in their sentiment scores (minors mean=-1.1; adults mean=-1.9). The most negative (overall) averaged sentiment score for the minors was -26.28, and -14.50 for adults. The most positive (overall) average sentiment score for the minors was 2.85, and 4.63 for adults. On average, minors vary more in their sentiment (minors range=29.14; adults range=19.14), presenting a wider range of emotions, when compared to adults. The difference that was found to be statistically significant at the .05 level.
Table 2. Descriptive statistics for overall sentiment—minors and adults
Note: Independent Samples Mann-Whitney U Test used (null hypothesis rejected).
For the minors, the before/after analysis tested whether there was a statistically significant difference between a minors sentiment score before they turned into an adult (before-19) compared to their sentiment score after they turn into an adult (after-19). For the adults, we divided the before and after as the first 50% of posts, compared to the last 50%. We also examined the first 50% versus the last 50% of posts made by minors to determine whether differences could also be found with adults when using this alternative definition.
Figure 1 display the overall (averaged) sentiment scores before and after the cut points described above. It can be seen that the overall sentiment scores for minors before turning 19 were significantly less negative than scores found after turning 19—suggesting an increase in negativity after transitioning into adulthood. Similarly, the first 50% of sentiment scores (overall) for adults were significantly less negative than those of the second 50%—suggesting an increase in negativity throughout their time on the forum.
Figure 1. Before/after and 50/50 trend (*p<.05)
Only 33 minors had both before and after 19 sentiment scores, as 15 individuals stopped posting before they entered into adulthood.1 The difference before (-1.16) and after (-2.9) turning 19 was not found to be statistically significant at the p<.05 level (p=.08) (see Table 3). A larger sample size may produce results closer to statistical significance. These results are suggestive of a difference, but the lack of significance warrants caution.
For the minors group, the first fifty percent of posts were averaged up and compared to the second fifty percent of posts, for all 48 users. As the variables were not normally distributed, a nonparametric test was employed to compare the sentiments expressed in the first and second half of each user’s posts (the related samples Wilcoxon signed rank test) – see Table 3. The same technique was used for the sample of adults (n=48). There was no statistically significant difference between the first 50 percent of sentiment scores (mean=-1.07) and the second 50 percent of sentiment scores (mean=-1.17) at the p<.05 level for the minors group, suggesting relative consistency in speech patterns, and sentiment over time. This finding reinforces the interpretation that entrance into adulthood may be a significant turning point for this sample of youth, enough to affect the nature of the sentiments they express online.
Table 3. Descriptive statistics and Wilcoxon signed rank test results—Minors and adults with 50/50 sentiment scores
Wilcoxon signed rank test
For adults, we did find a statistically significant difference at the p<.05 level between the first 50 percent of sentiment scores (mean=-1.50) and the second 50 percent of sentiment scores (mean=-2.29). The earlier posts of forum participants tended to be less negative than the most recent ones we examined.
Individuals displaying above-average negativity in their sentiment trends were extracted for further examination. In doing so, user sentiment trends were classified as being either consistent or inconsistent, with consistent trends being broken down into the following sub-groups: ‘consistent / negative,’ ‘consistent / positive,’ and ‘consistent / neutrality.’ Using qualitative analysis, we were able to further examine these trends by reading individual posts to confirm trend accuracy or provide alternate conclusions.
Of the 48, 25 minors were considered to possess inconsistent sentiment trends (outside the average group range of 4.95). Similarly, 23 minors were considered to possess consistent trends, with nine displaying positive overall linear trends, five displaying negative overall linear trends, and nine displaying neutral overall trends.
Of the 48, 22 adults were considered to possess inconsistent sentiment trends (outside the average group range of 8.32). Similarly, 22 adults were considered to possess consistent trends, with two displaying positive overall linear trends, seven displaying negative overall linear trends, and thirteen displaying neutral overall trends.
Minors were considered for further qualitative analysis if they met the following criteria: remained on the forum during transition into adulthood (n=33); presented either inconsistent trends, or consistently negative trends (n=30); displayed a change in sentiment (negative change) via their trajectory within 1 year before and 1 year after turning 19 (n=21). Thus, the final number of minors examined was 21. Adults were considered for further qualitative analysis if they met the following criteria: met the required amount of time periods to generate a range (n=44); presented either inconsistent trends, or consistently negative trends (n=29); displayed a change in sentiment (specifically negative change) via their trajectory within 1 year before and 1 year after their half-way mark on the forum (n=22). Thus, the final number of adults examined was 22.
Based on the quantitative findings of the current study, it was determined that the minors, on average, exhibited a statistically significant change in sentiment scores before and after transitioning into adulthood (at age 19). Similarly, the quantitative findings for adults revealed a statistically significant change in sentiment scores (on average) before and after the halfway mark of their time on the forum. While there did not appear to be any obvious consistencies regarding posting behavior (frequency of posts and sentiment of posts) and time spent on the forum surrounding this transition, two main themes were discovered nonetheless. The following section will contextualize and further elaborate on the possible impact of this turning point for minors, as well as the development (or lack-there-of) of sentiment occurring online for the adults. The group of adults presented identical themes to that of the minors – with differing examples and numbers of individuals found in each theme.
Minors and adults who showed consistency in their narratives were the most negative, and sometimes violent in their speech, but did not develop into (or out of) this type of online behavior as they transitioned into adulthood. Instead, these individuals tended to remain constant in their narratives over time. Even though they were not necessarily “well-received” by others online—as their posts sometimes appeared to be too extreme for the majority of users to accept—these users did remain on the forum, discussing with others and networking (forming both negative and positive relationships with others).
While the sentiment trajectory for the following users was originally classified as being inconsistent, it is important to note that the peaks or drops in 6-month data points (given the trajectory classification of being inconsistent) occurred during sudden decreases in posting behavior. Thus, there would not have been enough posts to average a sentiment score reflective of a consistent trend. Furthermore, all the users were found to be predominantly negative in their sentiment trajectories.
While not the most negative or radical of the users examined, the following individuals displayed consistently negative sentiment scores throughout their time online (regardless of sentiment trends that might have suggested otherwise) as a result of discussing/debating issues pertaining to religion, political affairs in the Middle East and the West, and the variation of appropriate cultural norms between the Middle East and the West. A large portion of the most negative posts can be attributed to news articles that were copied and pasted into a comment—specifically discussing these three topics.
User M-11 was chosen for his consistent rhetoric regarding the nature of Al Qaeda, terrorism, and Islam. While he does not express support for extremism, his sentiment trend was consistently negative as a result of the topic discussed, and words used.
Before-19: I know that Islam doesn’t support terrorism. Al Qaeda are deranged, these are the same kind of people that would behead their daughters if they accidentally showed their face, you can’t expect rational thinking from them. If they really want to protect muslim lands, they should fight man to man with foreign troops on their lands, not fly overseas and blow up innocent people who have nothing to do with any of that [Edited Out] and just want to live their lives and make a living.
After-19: I think wahabis/salafis are a bigger enemy of Islam because they’re an enemy from within.
This adult user considered himself to be a "loose Salafi", frequently quoting the works of Anwar al-Awlaki, arguing with others online (most notably another minor sampled in the current study), and somewhat "security-aware." User A-33 was not necessarily extreme in his sentiment, as he was quite eloquent in his writings, but was clearly attempting to sway the minds of individuals online – most likely as a result of his age (45 at start – oldest in the sample). He started off confident, did not sway in his opinions, and eventually became an Ummah moderator. The following quotes represent various time points throughout his trajectory online—from start to finish:
My source is the Taliban themselves, the Mujahideen themselves, a 11 part Jihad lecture which talk about everything by Anwar Al Awlaki.
Hereafter Series by Anwar Al Awlaki. I recommend every muslim give it a listen. Life and times of Umar ibn khattab by anwar al awlaki is a great series.
I respect Anwar Al Awlaki rahimullah because he spoke about it heavily then actually did it [referring to hijra, or the migration or journey of the Islamic prophet Muhammad and his followers from Mecca to Medina].
Re: security awareness: I don’t think anyone should be putting up contact information up on public forum.
The subsequent examples are of users (one minor and one adult) who displayed extreme, even radical, posting behavior from entrance into the forum—throughout the transition into adulthood, and past the halfway mark —until exiting the forum.
Even though User M-44 began on the forum at age 18, the majority of his posts could be found during this before-19 phase. It seems as if this particular user differed from all the rest in the sense that he immediately began posting (and engaging in) discussion topics of a more negative, even radical, nature from the moment he joined the forum. User M-44 acted as an "instigator," posting extremely negative, even violent statements. User M-44 appeared to be fanatical from day one on the forum – character traits which did not develop over time. He remained extremely negative until his last post on the forum (in 2011, 8 years). He also exhibited the traits of superiority and unwavering confidence in his person, religion, beliefs, and culture:
We Shias are superior muslims … and we are knowledgeable and mature. In Shiaism we think that religion comes first and then world. Well you Christians & Jews aren’t even true with ur religion. drinking alcohol and eating pigs makes u mean and u cant think out of ur world ... We are waiting for the mesiah whose name is Mehdi and he will come and inshallah destroy Satan America & Satan Israel ... Inshallah he will behead u MARINE BASTA*DS one day.
These types of users were not well received by others, especially User M-44. As such, responses from others on the forum were of concern and sometimes fear. Many of these responses could be attributed to the extremely negative, and often violent personal poems and chants that User M-44 wrote and posted on the forum. These poems and chants consisted of topics pertaining to martyrdom, killing, vengeance, hatred, terrorism, and the “Great Satans”: America and Israel.
From the beginning of his time online, User M-44 stated his fervent beliefs in Islam and what he considered to be "acceptable" religious/cultural behaviors that one should partake in if they consider themselves a "follower." These activities/behaviors consist of, but are not limited to: refraining from listening to music; practicing Zanjeer – hitting oneself on the back with a chain of blades (Zanjeerain); living in an Islamic state; getting married before age 18; being a pious wife (for women); and acquiring military training (for male Shia Muslims).
I ordered the Zanjeerain which were made on speacial order and…they came through regular mail from pakistan and it didn’t get caught in customs. I live in Canada and it is very difficut to do zanjeer matam here.
Salam, if anyone commit these heineous crimes so he should be handed to death or beheading squadAnywayz the solution to this is to impose Islamic Govt (not puppet one). Get children married when they are 18.
I think every shia must have a Military Training ... They must know how to use the gun … I know how to use Ak47 and 9 MM.I am planning to experience grenade throwing, bottle bomb throwing, rpg firing soon when i go to pakistan next year.
In addition to stating these behaviors as a responsibility of all Muslims following Allah, User M-44 wrote his articles supporting physical action, posting them to the forum, in an attempt to motivate others to become inspired towards the Jihad. Some of these articles received fairly positive responses of encouragement and support.
The reason our Imam is disappointed from us is because we are just replying to his call of help emotionally but not physically.
The conclusion of this article is we should participate in protests which are against the US, Israel, Britain and there Allies. This is the only way i see that we can answer Imam Hussain call emotionally and practically.
User M-44 displayed a fervent and unwavering desire to become a martyr, and for others to learn how they too could engage in martyrdom. Throughout the users’ time on the forum, this theme of a passion for martyrdom was the most constant. As can be seen in the proceeding sub-section for User M-44, there is supposedly a long line of martyrs in his family (back in Pakistan)—hence his desire to be a martyr.
Salam, My one wish that i wish come true is that i get shahadat or martydom ... I wanna get extreme martydom which is that my body gets into so many pieces that no one could recognize me … I know people will now call me extremist ... If u really wanna know how u can get shahadat so plz read the biography of Allama Shaheed Arif Hussain Al Hussaini and Shaheed Dr. Mohd Ali Rahimi Newayz i am not afraid of death as i am also the member of the gusul-e-mayyut commitee I think every buddy should learn how to give ghusul as it will get ur death fear out of u.
Death is sweeter than honey … Love Pakistan & Willing to die for Pakistan.
The second adult individual found within theme one presented himself in a manner that alerted authorities in the United States to his online activities. User A-44 was negative, even violent in his sentiment from the very beginning until he was arrested by the Federal Bureau of Investigation (FBI) on terrorism charges in 2014. This individual attempted to motivate others towards radical ways of perceiving Islam and was referred to in a recent mainstream media article as helping change the way young Muslims radicalize online.2 User A-44 was an avid Anwar al-Awlaki poster and displayed an unwavering (and consistent) belief that all Jews needed to be eradicated.
Dear brothers and sisters, it is as clear as day and night that the Jews are the enemy of mankind and even the environment. It is crystal clear that the Jews (may Allah swt destroy them) run the media, banking, USA foreign policy, and the porn industry, etc. Brothers and sisters rather than being kept busy refuting this bacteria we should focus on ridding Judaism from this earth.
In addition, User A-44 openly expressed his support for organizations considered to be terrorist groups by the Canadian government (CC section 83.05), even providing advice to others online on how to travel overseas and partake in activities alongside these groups. User A-44 openly referenced the importance of online forums—particularly the one he took part in—for known extremists and terrorists, and continually attempted to recruit and inspire others online:
We call you to Islam and to leave the lie that is Wall Street. Come to REAL ISLAM in its totality and make the world a better place. ISLAM IS THE ONLY SOLUTION!!!
It should be noted that User A-44 was arrested as a result of his activities on YouTube as well as a personal Muslim extremist website he co-founded with several other individuals (all of whom were arrested as well).
Some of the minors and adults considered to be relatively negative in their speech evolved into more extreme forms of posting as they transitioned into adulthood (minors) or passed the halfway mark (adults). There was an overall finding of increased frustration and annoyance at other users during the latter periods of their time online. Such a finding is consistent with the earlier quantitative findings, which revealed there to be a statistically significant difference between the first 50% of posts and second 50% of posts (for adults), and marginal significance for the sample of minors displaying more negative emotions in their online posts as they entered into adulthood.
The changes observed for the least negative users could be explained by individuals developing general life interests, but not necessarily towards more negative (or even extreme) opinions/beliefs, as well as overall irritability with one another. In other words, their opinions did not change, but the topics they chose to engage in did – hence the increasing negative sentiment scores. The following examples illustrate the development, which occurs over time (specifically during the transition into adulthood, and the immediate entrance into the latter half of posting), for the pseudo-extreme individuals:
User M-19 seemed to be posting many more discussion points and articles on conspiracy theories, Iran, terrorism and suicide bombers immediately preceding his transition into adulthood. Such a change in sentiment can be visually depicted in Figure 2, during the time period of analysis 2005(1).
All this anti-Iran propaganda they’re publishing everywhere is an attempt to turn the public against Iran so that if America tried to take some sort of action against Iran (because of nukes, the strength of the Islamic Govt. etc.) then there would hopefully be less critical public opinion.
There are more useful ways to do Jihad than in a testosterone fueled machomanegotistical attempt to satisfy one's arrogance by going to Iran and playing with your toy guns. If I was to take part in ''physical jihad'', it would be in defense of Iraq (or Ireland). Jihad is for Islam not for a country.
Figure 2. User M-19—Example of increased negativity (after 19)
Furthermore, User M-19 begins reading and studying more on Islamic history, specifically martyrdom. This change in topic (and therefore sentiment) can be seen in Figure 2, during the time period of analysis 2005(2).
I think the day of this great scholar’s martyrdom should be a message for all of us. It has now become clear that the government can not protect us. And that our silence and patience gives more courage to the SSP. I believe this is the time when we must stand up for our rights, and fight back, on all fronts.
Innalilahi Wa Innailaihi Rajioon. Allah has chosen him to be a martyr on the hands of His Prophet’s killers who know that we have a long battle with them and the blood of martyrs especially our leaders has always taken our resistance to a higher and stronger level.
While User A-29 began his time online discussing political, cultural, and other "sensitive" subjects (such as the Holocaust), he then developed from expressing himself in a neutral manner to expressing himself with more negative statements – directed towards Middle Eastern rulers who were in power and “wreaking havoc on their people” (as expressed by User A-29):
Salaam, theirs peace in the north and the south isn't really that stable plus the shabab have killed innocent people, I'm all for shariah law but you can't kill innocent people to get what you want.
May Allah destroy bashar al-kalb and raise him with his father hafez on qiyamah. Ameen, and may Allah destroy the raafidah.
I make dua every night for Allah to destroy him. He's a shaitan, who kills my fellow Brothers and Sisters.
User A-29 then goes on to express his newfound interest in learning from Anwar al-Awlaki:
Instead I listen to lectures (Anwar Al-Awlaki, Ali Timimi etc) so try and incorporate that. May Allah make it easy for you. Ameen.
Alternatively, the remainder of users found in theme two displayed signs of development as a result of engaging in life experiences (which were reflective online) but were not considered "extreme" in nature—a finding which should be expected for the majority of the online community. These individuals exhibited a slight development of frustration and annoyance directed towards other individuals on the forum as time went on.
User M-22 became involved in heavier and long/intense religious debates (and general debates) after transitioning into adulthood, as opposed to simply posting articles – as he did before adulthood (examples of these articles will not be provided for the sake of space). User M-22 began making comments in reference to his new interest of learning about martyrdom – however, not in an “extreme” or worrisome manner. In addition, new opinions and interests appeared to be forming with regards to political / cultural issues in the West versus Pakistan, interests which attracted the attention of forum moderators who consequently deleted his post for being too offensive.
User M-22 expressed he had not been going on the forum as much lately. When he did, he would write long messages (fairly well reasoned). Some posts displayed hints of frustration and annoyance when he was forced into writing a long response.
First of all [user name edited out] I suggest you quit with your lame remarks and your holier than thou attitude. You are a no one. I don't know why you seem to have an interest in my personal life and what I do (wasn't it you who randomly brought up some story about me going to school on Ashura on some next thread :lol:) and what sites I visit or don't visit. Do you have some sort of a Website-tracker installed on your computer that tells you what websites I visit for nohas?
User A-29 increasingly developed into frustration over his time online – as represented by the following comment made in the latter half of his online trajectory. This is one of the last posts made by User A-29:
I'm done trying to justify and defending some of you who bring this website into disrepute, so by avoiding it i've removed it completely for the time being until further notice to save the headache, really we don't need to be dealing with such issues now All of those threads should be removed.
As a result of being immersed in an online community whereby more intense and even “extreme” political and cultural topics are discussed, such a development into negativity (based on sentiment scores) seems fairly reasonable. Throughout time spent on the forum, it would be expected that individuals would eventually begin discussing topics of a more negative (possibly extreme) nature, as this would constitute a natural progression throughout the forum based on developing interests and exploration into the various sub-forums and threads available to a user.
This section compares the top five most negative individuals (as per their sentiment scores and trajectories) with the qualitative findings to determine whether or not those most negative quantitatively are also those most negative qualitatively. These individuals were drawn from both the minors group and the adults group. To identify these users, SentiStrength scores were averaged up per user to determine overall sentiment scores. Those users with the five most negative overall sentiment scores were selected for comparison with the top five users qualitatively determined to be most negative/violent. Table 4 displays the top five most negative individuals as per the SentiStrength results, and the qualitative analysis findings – separated.
Table 4. Top 5 most negative users—SentiStrength vs. qualitative analysis
Ave. sentiment score
All five users found to be most negative qualitatively were determined by the themes in which they were classified. All individuals found in theme one (n=5) presented as either extreme or pseudo-extreme, and as such, were classified as the top five most negative users qualitatively.
As can be seen from Table 4, the top five most negative users as determined by SentiStrength did not appear to be the top five most negative users as determined through qualitative analysis. The sole case that may have constituted an exception would be that of user A-9, who was determined to be quantitatively one of the top five most negative individuals. However, he was not considered to be one of the top five most negative individuals qualitatively (he would, however, constitute a top 10 member).
The findings of this analysis support the overarching argument (of the current study) that qualitative analysis is required to better understand and contextualize the quantitative findings of SentiStrength. These two types of analyses would appear to supplement each other, providing further support for the notion of a mixed-method approach when examining online trends of sentiment over time.
The current study set out to examine the development of negativity in the online context, with a specific focus on minors engaging in online Islamic forums. We used theoretical guidance from Life Course Theory (LCT) and General Strain Theory (GST), to determine if certain turning points in life could affect an individual's online sentiment trajectory. Quantitative results suggested the transition into adulthood (for minors) to be important, as this point in time was reflected by a statistically significant increase in negativity (compared to their pre-19 sentiment scores), and a non-significant finding when comparing first 50 percent to second 50 percent of posts by minors – reinforcing the interpretation that entrance into adulthood may be a significant turning point.
Sampson and Laub (1990) stipulate that social bonds developed in adulthood seem to have a fairly significant effect on modifying (or reducing) criminal and deviant behaviour developed during adolescence. While youth tend to age out of certain life behaviors and mentalities (as they mature), the same cannot necessarily be said for that of negativity (or negative affect). The current study determined negativity online to increase after transitioning into adulthood. The current findings do not seem to support the notion that youth will develop out of (or reduce) the amount of negative affect they experience and express online. Additionally, analysis of the adults group further supports the idea that negative affect does not necessarily diminish over time. These findings lend support to previous literature on affective development whereby negative affect increased with age (Ferring and Filipp, 1995). However, these studies were conducted in the offline setting, whereas the current study analyzed sentiment in the online setting – a factor that would require further research and analysis in order to understand the impact of such a contextual change, if any.
While most users displayed a change in sentiment (quantitatively) and overall posting content (qualitatively) before and after transitioning into an adult (age 19), for a select few, there did not appear to be any evidence of changes during this time. Through qualitatively analyzing these individuals, two overarching themes were discovered for both the minors and adults group. The first was that the most negative/radical individuals tended to be consistent in their narratives over time. Initially, it was not well understood the extent to which the more extreme individuals would be acting as "inspirers," "motivators," or "indoctrinators" towards others online. Thus, it became important to understand the effect certain (more violent) individuals had on others online – particularly youth. Essentially, these users did not appear to “sway” others with their thoughts and manner of dialogue. In fact, the more infuriated they became, the less they were able to get their point across, as others were either ignoring them or refuting their arguments. Furthermore, the majority of minors examined appeared to be quite resistant to, and able to identify, other users of an extreme negative and even radical nature.
Of those displaying extremely negative online sentiment trajectories, the lack of changes in sentiment suggests they may not have been affected by any significant turning points during this time. It is possible, however, that these individuals experienced life-altering turning points offline (before entering into their forums). User M-44, for example, revealed viewing the violent death of several family members (through acts of martyrdom) back in Pakistan. Such an environment of violence and extreme emotional distress may have stunted certain aspects of this individual’s development. Ginwright and James (2002) refer to the more precarious and detrimental life conditions as “social toxins” (p.28) – otherwise understood as representing “the degree to which the social world has become poisonous to a person’s well-being” (p.29), with specific reference to extreme forms of violence, brutality, and threats to the family. Furthermore, these “social toxins” create an unhealthy environment in which youth must live and mature, preventing them from adequately (or normally) developing (Ginwright and James, 2002; Brooks-Gunn, Ducan, Klebanov, and Sealand, 1997).
As was stated numerous times throughout user M-44's online trajectory, he seemed to display a desire to become a martyr himself (following in the footsteps of previous family members). Many academics to date believe oppressive and toxic environments foster emotions and desires of "depression, loneliness, and suicidal tendencies" (Ginwright and James, 2002, p.31; Brooks-Gunn et al., 1997). Thus, the significance of growing up in an environment of violence and trauma could very well stunt (or prevent) a youth from developing in a healthy manner, which may come across as a lack of development – as portrayed in the online setting.
Furthermore, LCT would stipulate that the “timing and sequencing” (Sampson and Laub, 1993, p.254) of turning points is of vital importance when examining the effect they have on later behavior. According to Agnew’s (1992) GST, anger is a vitally important emotion and is considered to be a key factor in the development of criminal behavior (Froggio, 2007). With such traumatic events occurring so early on in life, it is possible to foresee how the frustration and anger experienced offline could develop into the violent (even radical) behaviour seen online.
The second theme was ‘evolution in narrative (with pseudo-extreme and least negative individuals).' These individuals tended to vary in their speech patterns over time, with the turning point of entrance into adulthood/university (or the halfway point for adults) displaying a noticeable change in each user's sentiment and content of posting. The majority of minors, however, did not necessarily develop into more negative and extreme forms of narrative. They simply portrayed a development of maturity and responsibility while becoming increasingly frustrated with other users online.
The individuals who developed into more extreme forms of negativity merely increased their use of negative and extreme words (and posting of extremely negative news articles). They did not necessarily develop into more extreme forms of thinking or acting. For both groups, it was common for users to develop within their forum, increasing their interests in more controversial topics pertaining to politics, international affairs, and terrorism in the Middle East.
Few academics have addressed the notion of online peer interactions and communications leading to frustration and annoyance with one another (McInnerney and Roberts, 2004). Huffaker (2010) notes the impactful nature of negative affect (or negative emotions) expressed online in encouraging further negative feedback and reciprocity. Thus, it would seem logical to assume that the development of frustration over time within this group of individuals (in theme two) could very well be attributed, at least partially, to the somewhat controversial setting in which these conversations took place.
According to Sampson and Laub (1993), the ability for a minor to transition seamlessly into adulthood becomes a vital factor for healthy development. Based on the quantitative findings of the current study, this transition into adulthood did coincide with an increase in negative sentiments. While it is difficult to conclude with a high degree of accuracy that the increase in negativity can predominantly be attributed to entrance into adulthood, the results do suggest there to be some level of relationship between these two variables. Qualitatively, it became evident that almost all of the minors partaking in discussions online were experiencing stress and strain during this time as a result of more general life factors (such as leaving home, starting university, attaining a job, etc.). In order to cope with these stresses, many began using their respective forum as a stress outlet and means of support. As such, it would appear as if the majority of minors (and even adults) found solace in their new online community, easing their transition into adulthood.
With the exception of topics pertaining to high school (and related youth interests), both minors and adults tended to discuss fairly similar subject matters in their respective online forums: issues of politics, culture, and terrorism (specifically in the Middle East). Additionally, the traits of rationality, objectivity, and maturity all seemed to be present in individuals of varying ages, suggesting these traits to be more so attributed to the individual themselves, as opposed to traits such as age and gender. Where the groups did differ was in their manner of speech and additional topics of discussion. It became evident that the minors were more naive in both their writing styles and opinions, whereas the adults presented themselves in a more logical and even rational manner (with a few exceptions).
This study was unable to examine motives for participating in discussion forums. It is reasonable to expect that motives to enter the forum as an adolescent could be different than the ones found for adults. Some studies have found participation in online community groups to be the result of social identification – or one’s attempt to establish and identify themselves in the social setting via exposure to others online (see Ducol et al., 2015 for a review). Thus, demographics and rationales of participation in these forums would be an appropriate avenue for future research, as such an analysis may shed light upon the differences between minors and adults found in the current study.
Many forum users were frequently asking questions with regards to the religion of Islam and Islamic culture; these individuals tended to be the ones who were curious and seeking knowledge – especially if they were ‘converts' to Islam. While this in itself is not associated with violent extremism (as many seek further knowledge to understand their religion better), it has been suggested by scholars, intelligence professionals, and even former terrorists, that the act of seeking knowledge can highlight individuals for recruitment by terrorist organizations (Masi, 2014). Mubin Shaikh – a former recruiter (in Toronto) for the Taliban, now acting as a CSIS operative – stated in an interview with Alessandria Masi (2014) of International Business Times, that when recruiting:
There [are] certain things looked for: people who didn’t know the religion as much, [and] people who were converts, because converts would probably have problems with their parents at home, so they were more likely to stay in our [terrorist organization’s] company (para. 4).
While the current study did not set out to profile terrorists, or even potential terrorists – as such a study would require a different data set and analytic strategy –there were a select few users who self-reported actions of terrorism, and some who self-reported intent to engage in actions of terrorism. Although these individuals were the minority (in both the minors and adults groups), they were still present within the relatively small sample of users collected, and extracted, for analysis.
While graphing sentiment trajectories provided a good base from which to sample inconsistent or consistently negative online trends, once qualitatively analyzed, it was determined that these sentiment trends did not always correspond to the qualitative findings – qualitatively consistent sentiment profiles may have been quantitatively inconsistent. Some reasons as to why this may be are that 1) several time periods of analysis (6-months) had only a mere few posts from which to analyze – skewing the average for the time period of analysis, and creating an inconsistent trend; 2) Sentiment analysis captured articles that individuals had posted online – sentiment which may have been negative, but not necessarily representative of the individuals personal opinions/beliefs, and 3) in some instances, sentiment analysis captured the original post for which said user was responding. Such findings suggest more work is needed to develop the SentiStrength program better and obtain more accurate and representative sentiment scores, as well as to better understand how sentiment analysis can provide accurate and supportive findings to corroborate the qualitative (or “contextual”) aspect.
Both approaches, however, did measure negativity over time, allowing for an overlap in objectives. Each analytic tool taps into a separate and distinct concept, as both measure their separate phenomenon over time – with SentiStrength measuring user sentiment from a quantitative approach, and qualitative analysis accounting for the contextual factors surrounding user sentiment. It is difficult to determine the exact amount (or percentage) each tool explains or accounts for in the other (how much they overlap), but when comparing the top five most negative users as determined quantitatively, with those determined qualitatively, there was no overlap between the two groups, suggesting a limited amount of overlap between the two analytic tools.
As the current study attempted to address a somewhat under-researched area of study, several limitations should be noted. First, only the negative individuals displaying inconsistent or consistently negative sentiment trajectories were examined qualitatively. As the current study intended to focus predominantly on the development of negativity over time, such a research decision to focus on negative trajectories was warranted, but does constitute a bias and limitation. Second, the inclusion of only two turning points (entrance into adulthood, and entrance into university) may present a limitation of the current study, as the inclusion of additional turning points—stipulated through LCT literature—may have explained more. For example, once the adults group passed the halfway mark of their time online, they became significantly more negative over time, suggesting there could be other factors (and turning points) effecting the progression/development of sentiment online.
Third, sentiment analysis (or SentiStrength) tended to capture articles in their entirety, which individuals had posted online as part of their comment or response. As such, the overall sentiment score generated for those posts would have been somewhat unrepresentative of the user's true opinions, thoughts, and beliefs – especially if the article was being posted as a joke or counter-argument to the point they were making. Likewise, in some instances, sentiment analysis captured the original post a user was responding to – meaning there would have been a sentiment score generated for a post that technically encompassed another user’s sentiment (or opinions and beliefs), thereby skewing the overall sentiment trajectory for that user. Lastly, the SentiStrength program was not designed to account for, or capture, the contextual elements of sentiment online. Individuals could have been using sarcasm, and SentiStrength would have generated an overall score reflecting the positive or negative nature of said comment. In addition, users may have been positively agreeing with a very negative (or even violent) statement – warranting a negative overall score – yet, SentiStrength would have generated a positive score based on the words being used (and vice versa).
Overall, literature to date has failed to adequately address within-individual development in the online context. The current study sought to address this limitation by doing exactly that. While the findings of this study should not be considered all-encompassing – to the extent of explaining (and understanding) the exact process by which the Internet serves to foster personal development – use of the Internet (specifically online forums) did appear to serve as a strategy to understand the development of maturity online. Qualitative findings presented ample examples of such a concept (most of which were present in theme two), with only a select few (those of a more extreme and radical nature) failing to exhibit this trend.
Future research should examine the impact of major events on the sentiments expressed by users. For example, determining whether these major events have an impact on individual user sentiments, beyond the larger tendencies captured for these individuals over time. Such an endeavor would prove useful for understanding the impact of external events on the internal development of each user (as expressed via their posts and associated sentiment scores).
While it would have been ideal to use the current study (and subsequent findings) to inform the process of radicalization in the online setting, the current design was not intended to accomplish such a goal. Future research, however, might benefit from expanding the research design to include a broader sample of all nationalities and several more Islamic forums (e.g. Scrivens, Davies, and Frank, 2018). Research goals would need to focus specifically on better understanding the process of internal (individual) development one would need to experience online in order to become radicalized—including analyzing turning points and identifying ‘indicators' of possible radicalization—as well as how and whether violent radicals may influence others online, if at all.
Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30(1), 47-87.
Agnew, R. (1997). Stability and change in crime over the life course: A strain theory explanation. In T. P. Thornberry (Ed.), Developmental theories of crime and delinquency, 101 – 132. New Brunswick, NJ: Transaction.
Berger, J. M. (2011). Jihad Joe: Americans who go to war in the name of Islam. Washington, D.C: Potomac Books, Inc.
Berkowitz, L. (1989). Frustration-aggression hypothesis: Examination and reformulation. Psychological Bulletin, 106(1), 59-73.
Berson, I. R., & Berson, M. J. (2005). Challenging Online Behaviors of Youth Findings from a Comparative Analysis of Young People in the United States and New Zealand. Social Science Computer Review, 23(1), 29-38.
Brooks-Gunn, J., Ducan, G. J., Klebanov, P., & Sealand, N. (1997). Do neighborhoods influence child and adolescent development? American Journal of Sociology, 99(2), 353–395.
Cassell, J., Huffaker, D., Tversky, D., & Ferriman, K. (2006). The language of online leadership: Gender and youth engagement on the Internet. Developmental Psychology, 42(3), 436-449.
Cernkovich, S. A., & Giordano, P. C. (2001). Stability and change in antisocial behavior: The transition from adolescence to early adulthood. Criminology, 39, 371-410.
Cohen, E., & Goldschmidt, D. (2015, December 21). Ex-terrorist explains how to fight ISIS online. CNN. Retrieved on January 27, 2016. Retrieved from http://www.cnn.com/2015/12/18/health/al-quaeda-recruiter-fight-isis-online/
Dollard, J., Doob, L., Miller, N., Mowrer, O., & Sears, R. (1939). Frustration and Aggression. New Haven, CT: Yale University Press.
Donker, A. G., Smeenk, W. H., & Van der Laan, P. H. (2003). Individual stability of antisocial behavior from childhood to adulthood: Testing the stability postulate of Moffitt's developmental theory. Criminology, 41(3), 593-609.
Ducol, B., Bouchard, M., Davies, G., Ouellet, M., & Neudecker, C. (2015, June 29). Assessment of the state of knowledge: Connections between research on the social psychology of the Internet and violent extremism. Public Safety Canada.
Elder, G. H. (1985). Life course dynamics: trajectories and transitions, 1968-1980. Jr, Social Science Research Council (U.S.). Ithaca: Cornell University Press.
Ferring, D., & Filipp, S. H. (1995). The structure of subjective well-being in the elderly: A test of different models by structural equation modeling. European Journal of Psychological Assessment, 11, 32.
Fredrickson, B. L. (2004). The broaden-and-build theory of positive emotions. The Royal Society, 359, 1367-1377.
Froggio, G. (2007). Strain and juvenile delinquency: A critical review of Agnew's general strain theory. Journal of Loss and Trauma, 12(4), 383-418.
Ginwright, S., & James, T. (2002). From assets to agents of change: Social justice, organizing, and youth development. New directions for youth development, 2002(96), 27-46.
Guan, S. S. A., & Subrahmanyam, K. (2009). Youth Internet use: risks and opportunities. Current opinion in Psychiatry, 22(4), 351-356.
Hsieh, H. F., & Shannon, S. E. (2005). Three approaches to qualitative content analysis. Qualitative health research, 15(9), 1277-1288.
Huffaker, D. (2010). Dimensions of leadership and social influence in online communities. Human Communication Research, 36(4), 593-617.
Kennedy, H. (2012). Perspectives on sentiment analysis. Journal of Broadcasting & Electronic Media, 56(4), 435.
Loeber, R. (1990). Development and risk factors of juvenile antisocial behavior and delinquency. Clinical psychology review, 10(1), 1-41.
Masi, A. (2014, September 8). ISIS recruiting Westerners: How the ‘Islamic State’ goes after non-Muslims and recent converts in the West. Retrieved on January 27, 2016. Retrieved from http://www.ibtimes.com/isis-recruiting-westerners-how-islamic-state-goes-after-non-muslims-recent-converts-west-1680076
McInnerney, J. M., & Roberts, T. S. (2004). Online learning: Social interaction and the creation of a sense of community. Educational Technology & Society, 7(3), 73-81.
McKenna, K. Y. A., & Bargh, J. A. (2000). Plan 9 from cyberspace: The implications of the Internet for personality and social psychology. Personality and Social Psychology Review, 4(1), 57-75.
Mroczek, D. K., & Kolarz, C. M. (1998). The effect of age on positive and negative affect: a developmental perspective on happiness. Journal of personality and social psychology, 75(5), 1333.
Neumann, P. R. (2013). Options and strategies for countering online radicalization in the United States. Studies in Conflict and Terrorism, 36(6), 431.
Ryan, J. (2007). Countering militant Islamic radicalization on the Internet: A user-driven strategy to recover the web. Institute of European Affairs, 23-82.
Sampson, R. J., & Laub, J. H. (1990). Crime and deviance over the life course: The salience of adult social bonds. American Sociological Review, 55(5), 609-627.
Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and timing points through life. Cambridge, Mass.: Harvard University Press.
Scrivens, R., Davies, G., & Frank, R. (2018). Measuring the Evolution of Radical Right-Wing Posting Behaviors Online. Deviant Behavior. Online First: https://doi.org/10.1080/01639625.2018.1556994
Shapiro, L. A., & Margolin, G. (2014). Growing up wired: Social networking sites and adolescent psychosocial development. Clinical Child and Family Psychology Review, 17(1), 1-18.
Shek, D. T., & Yu, L. (2011). A review of validated youth prevention and positive youth development programs in Asia. International Journal of Adolescent Medicine and Health, 23(4), 317-324.
Smith, J., & Baltes, P. B. (1993). Differential psychological aging: Profiles of the old and very old. Ageing and Society, 13, 551-587.
Subrahmanyam, K., & Greenfield, P. (2008). Online communication and adolescent relationships. The future of children, 18(1), 119-146.
Ybarra, M. L., & Mitchell, K. J. (2005). Exposure to Internet pornography among children and adolescents: A national survey. Cyberpsychology & behavior, 8(5), 473-486.
Young, K. (1998). Caught in the net: How to recognize the signs of Internet addiction and a winning strategy for recovery. New York: Wiley.
Philippa Levey recently graduated with her MA from the School of Criminology, Simon Fraser University. She is interested in life-course trajectories, especially those that involve violent extremism.
Martin Bouchard is a Professor of Criminology at the Simon Fraser University. His work focuses on the organization and dynamics of illicit markets and on examining the impact of social networks in various criminal career outcomes. His current projects include an examination of the social structure of gang violence and its implications for understanding the dynamics of gang conflicts.