Online Activities and Extremist Attitudes in Adolescence: An Empirical Analysis with a Gender Differentiation

1.1 Radicalization and Extremism

Given that most radicalization researchers agree on defining radicalization as the process of becoming an extremist (Neumann, 2013), a clear definition of extremism is needed. Of the numerous definitions that exist in the scientific discourse, we will refer to the one from Beelmann (2020). According to Beelmann, »political, religious, and otherwise conceived extremism [is defined] as a significant deviation in attitudes and behavior from basic legal and political norms and values within a social system (society or state) that seek their (at least partial) abolition and replacement (Beelmann et al., 2017). The process by which such patterns of attitudes and action emerge individually and ontogenetically over the course of development can then be labeled radicalization« (Beelmann, 2020, 2).

1.2 Social Learning Theory

As a theoretical framework for the explanation of radicalization processes, we refer to the Social Learning Theory of Akers, which has previously been applied in this field successfully (Wolfowicz et al., 2020; Pauwels & Schils 2016). Based on the assumptions of differential association theory (Sutherland, 1939), differential reinforcement theory (Burgess & Akers, 1966) and social cognitive learning theory (Bandura, 1977), Akers (1977) formulated the criminological theory of social learning. Akers assumes that deviant attitudes and behaviors are learned through interaction. The concept of differential association explains the process of learning deviant attitudes and values through direct social interaction with others (Akers, 1998). Differential reinforcement involves the perception of the consequences of specific actions, which can be rewarding or punishing. When actions are rewarded by others, this reinforces the likelihood of the action happening again (Akers, 1998). The adoption of deviant behaviors previously observed in others is referred to as imitation. Here, the extent of adoption depends on the observed consequences. If the consequences are observed to be rewarding, this promotes the adoption of the behaviors. Imitation is particularly important for the initial acquisition and execution of new behaviors. Through differential association and imitation, definitions are learned and reinforced, i.e., individuals acquire attitudes toward certain values and classifications of behaviors as right or wrong (Akers, 1998). It is further postulated that social contact, especially with peers, has an influence on deviant behavior. The theory was initially intended to explain only deviant behavior, but a large number of studies have applied it to the field of extremism as well (Akins & Winfree, 2016; Becker, 2019; Goede & Lopez Trillo, 2020; Lafree et al., 2018; Pauwels & Schils, 2016; Wolfowicz et al., 2021). While the influence of family on extremist attitudes and behaviors has been studied less frequently than that of radical peers, it has nevertheless been confirmed (Lafree et al., 2018; Becker, 2019). The Internet and especially social media can, in the sense of the theory, be seen as a place for socialization and social learning, since differential association as well as differential reinforcement can be experienced via social media.

1.3 Radicalization in Adolescence

During adolescence, individuals begin to search for their own identities, thus making them susceptible to new attitudes and more likely to join new peers. Adolescents are considered a particularly vulnerable group to engage in radicalization processes because they begin to seek belonging, recognition, and identity (Greve, 2007). They also seek adventure, thrills, and provocation (Benslama, 2017). In addition, the relationship between parents and children changes as adolescents. Some risk factors for radicalization, such as identity problems and other psycho-social aspects, play a particularly important role during adolescence (Benslama, 2017; Emmelkamp et al., 2020). They can be caused by the biological and psychological development of adolescents and are thus common in this age group. Adolescence is also an important period for political socialization (Beelmann, 2020; Niemi & Hepburn, 2010; Torney-Purta, 2004; Watts, 1999) and the development of attitudes that generally remain stable over one's lifetime (Sears, 1983, 1990). Young people initially experiment with different political positions, which encourages both the adoption of extreme political positions and short-term changes in basic positions (Alwin & Krosnick, 1991; Rekker et al., 2015). Extreme political positions and values are also considered an early feature of individual radicalization (Bliesener et al., 2021). In general, extremist attitudes are widespread among adolescents, as numerous studies from different countries have shown (Baier et al., 2019; Cherney et al., 2020; Krieg et al., 2022; Krieg et al., 2018; Muxel, 2020; Pfundmair et al., 2020; Pauwels & Waele, 2014).

1.4 Online Radicalization

One risk factor for radicalization that is often mentioned is the Internet. There is a consensus that the Internet in general cannot be a causal factor, but several types of Internet use can be considered relevant risk factors among others (Hohnstein & Glaser, 2017; Ducol et al., 2016; Behr et al., 2013). In other words, the Internet is not the only reason for radicalization. Instead, functions of the Internet can be described as catalysts that facilitate and speed up the process of radicalization (Pauwels et al., 2014).

Following the emergence of online communication, radicalization processes shifted from face-to-face interaction to online platforms (Koehler, 2014). Furthermore, the Internet is used to find answers to political, theological, or social questions (Holbrook, 2015). Hussain and Saltman (2014) point out the role of search engines in finding extremist content and the lack of counter-speech sites supporting moderate views. For individuals without prior knowledge of extremism, it is difficult to evaluate whether content is extremist or not (Hussain & Saltman, 2014).

People who have already been radicalized use the Internet to seek practical and ideological support (Leimbach, 2017). Besides websites and forums, such support is particularly found in social media (Guhl et al., 2020). Online magazines, such as the »Inspire« magazine of Al-Qaida, can also influence radicalization processes. Reasons for Internet use by extremist organizations include propaganda, recruitment, and funding, as well as logistics and planning (Winter et al., 2020). For such groups, the Internet has various functions, including networking, swapping of ideological ideas, disseminating extremist music, building an ideological worldview that is isolated from the »outside«, destroying the credibility of the mass media, and establishing an enemy image of the state and the media (Neumann, 2015).

However, most researchers do not provide specific definitions of ‘using the Internet’ or ‘being online’. The Internet offers a variety of functions and activities. The Internet also has cognitive (gaining knowledge) and social (communication with like-minded peers) incentives (Ducol et al., 2016). To prevent radicalization processes, it is important to know which online activities influence which kind of people and to what extent. The specific link between online activities and radicalization processes is unclear (Jukschat & Kudlacek, 2018). »Today’s Internet does not simply allow for the dissemination and consumption of ‘extremist material’ in a one-way broadcast from producer to consumer, but also high levels of online social interaction around this material. It is precisely the functionalities of the social Web that causes many scholars, policymakers, and others to believe that the Internet is playing a significant role in contemporary radicalization processes« (Conway, 2017, 80).

1.5 Empirical Findings

Pauwels and Schils (2016) examine the link between extremist content in new social media and self-reported political violence by distinguishing between active and passive forms. In their analysis of data from surveys with a total of 6,020 young Belgian people, they find that actively seeking extremist contacts online, participating in discussions with extremist content, and consuming more extremist online content are all significantly positively related to politically motivated violence against property. Similar results could also be obtained regarding politically motivated violence against people. Considering other variables, such as impulsivity, juvenile delinquency, or religious authoritarianism, they showed that contacts in the real world are just as significant as online contacts (Pauwels & Schils, 2016). Moreover, using the same sample, Pauwels and Hardyns (2018) find that active exposure to social media in terms of communication about extremist content reinforces the already positive effect of endorsement of extremism on self-reported political aggression.

The search for extremist online material was also analyzed by Frissen (2021). He analyzes the relationship between seeking online extremist material and cognitive radicalization among young Muslims in Belgium. He finds that consumption of beheading videos was sought by 36.11 % of the convenience sample, which is a relatively large amount compared to the share of participants consuming other materials, such as jihadist magazines or content in Facebook groups. However, beheading videos have only a small effect on cognitive radicalization and this type of material was the least associated in the model (Frissen, 2021).

Posting, as an active form of online behavior and general communication in right-wing online forums, has been analyzed by Scrivens, Wojciechowski et al. (2021) and Scrivens, Osuna et al. (2021), who find differences between violent and non-violent right-wing extremists. They find more posting activity among non-violent extremists in online forums (Scrivens, Wojciechowski et al., 2021). The authors argue that this may be explained by the clandestine character of violent extremists’ online behavior due to fear of law enforcement agencies. Wolfowicz et al. (2022) included four experimental and 49 observational studies with media-related risk factors for cognitive and behavioral radicalization in their systematic review. According to their review, passive and active forms of engagement with radical content on the Internet reveal small but potentially significant correlations with cognitive radicalization (Wolfowicz et al., 2022).

The extent to which the influence of the Internet on radicalization depends on gender has largely been ignored. Sanchez (2014) notes that the Internet has an influence on women who live in traditional societies. For example, the Internet allows girls and women to interact with strangers outside of the family’s control. Although much is known about gender dynamics in the radicalization processes with reference to different phenomena, only a few studies have explored female usage of the Internet in extremist contexts (Winter et al., 2020). For boys, the misogynistic ›Incel‹ culture shows an example of how gender-specific online activities manifest themselves with close ties to right-wing extremism (O’Malley et al., 2020; Pelzer et al., 2021). Since studies show differences between men and women in the extent of extremist attitudes and online behavior (Almenayes et al., 2014; Krieg et al., 2022; Decker et al., 2022; Goede & Lopez Trillo, 2020; JIM Study, 2021; Küpper et al., 2021), a more in-depth look at online radicalization with a gender differentiation is needed.

Methodologically, the body of literature on online radicalization contains a gap between a vast number of qualitative studies and only a few quantitative studies. Qualitative studies mostly focus on content that is published in extremist forums and social media (Gill, 2007; Gill et al., 2017; Jukschat & Kudlacek, 2017; Meier et al., 2022; Struck, 2019; Struck et al., 2020; Harrendorf et al., 2020; Mischler et al., 2019; Holbrook, 2015). Only a few attempts have been made to analyze these behaviors using a large quantitative sample, such as survey data (Gill et al., 2017; Wolfowicz et al., 2022). Some studies only include single case studies or anecdotes (Gill et al., 2015; Pearson, 2015).

1.6 Hypotheses

This study specifically analyses the link between different online activities and extremist attitudes in the field of Islamism and right-wing extremism among ninth graders in Germany. Based on Social Learning Theory (Akers, 1977), we expect that there is an association. Young people learn extremist views on the Internet through direct contact with extremists or through disseminated propaganda and subsequently develop extremist views themselves.

Because current research lacks a differentiation of various online activities, this study highlights those areas that we believe may be related to extremism. These activities can be active or passive. In particular, we focus on the consumption of political or religious content (passive behavior), and on the posting of political or religious content via social media (active behavior). Furthermore, we consider networking via social media as an important aspect, since previous research only included mostly (deviant) offline peers or friends (Wolfowicz et al., 2020). Other online activities such as gaming, learning, shopping, or streaming can theoretically be part of extremist activities, but are not the focus of this contribution. We assume that these activities must be distinguished because they are of different importance for extremist endeavors. While consumption is a key element of indoctrination, there is not necessarily a direct contact to extremist groups. It is possible that extremist content can be consumed and critically reflected upon.

Many radicalization models and theories assume that a radicalization process consists of multiple stages (Bliesener et al., 2021). Some note that the radicalization process begins with sensitivity (Doosje et al., 2016), a cognitive opening (Wiktorowicz, 2005), the urge to do something about injustice (Moghaddam, 2005), or moral outrage (Sagemann, 2008). Based on this, we deduce that there are also different stages during online radicalization. It is assumed that passive consumption is the first part of the radicalization process. Consumption could also include youthful experimental behavior that does not necessarily lead to radicalization. Only later on does it become clear whether consumption resulted from simple youthful interest or from the beginnings of radicalization.

Networking with people who already belong to a certain group or share certain dissenting views would represent a next stage in online radicalization and provide a space for social learning. Networking is also necessary for clandestine groups to disseminate and plan things that need to be kept away from the public.

Posting, on the other hand, is central to the spread of ideology in the digital age. For users, posting can be experienced as active participation in a social endeavor. For their audiences, who they often consider friends, these users can, intentionally or not, make their posted content seem trustworthy and disguise its extremist character. Only when extremist attitudes are hardened through consumption and networking does active posting occur. Thus, we see the active posting of extremist content as the latter part of online radicalization. Therefore, the specific link between different online activities and extremist attitudes can be analyzed.

The study also considers gender differences within the correlation between online activities and extremist attitudes. While in the offline world, Islamist women tend to be ascribed a passive role, studies have found that women take an active role on the Internet (Nisa, 2019; Nisa 2013). Since explicit Islamist propaganda on the Internet is also directed at girls and women, we assume that they are more active than boys, who are also active in the offline world. In the area of right-wing extremism, there is a classic gender role distribution in which males take the more active and dominant role. Therefore, we expect boys to be more active than girls in both the offline and online worlds. Following the Social Learning Theory, more activity leads to more extremist definitions, which can be referred to as extremist attitudes. This leads to the following Hypotheses:

Islamism:

Hypothesis A1: The more frequently young Muslims consume violent Islamist videos, the more likely they are to have Islamist attitudes.

Hypothesis A2: The more frequently young Muslims meet Muslim friends via the Internet, the more likely they are to have Islamist attitudes.

Hypothesis A3: The more frequently young Muslims post Islamic content on their social media profiles, the more likely they are to have Islamist attitudes.

Hypothesis A4: Muslims with Islamist attitudes use online activities with Islamic or violent Islamist content more frequently than young people with no Islamist attitudes.

Hypothesis A5: There are more young people who initially only passively consume Islamist content than young people who actively post. The number of young people who network online is somewhere in between.

Moreover, gender differences in attitudes and in Internet use in general lead to Hypothesis A6: In the field of Islamism, the use of the Internet plays a bigger role for girls than for boys.

Right-wing extremism:

Hypothesis B1: The more frequently young people visit websites with politically right-wing content, the more likely they are to have right-wing attitudes.

Hypothesis B2: The more frequently young people communicate with right-wing persons on the Internet, the more likely they are to have right-wing attitudes.

Hypothesis B3: The more frequently young people post things with right-wing political content on the Internet, the more likely they are to have right-wing attitudes.

Hypthesis B4: Adolescents with right-wing attitudes use online activities with right-wing content more frequently than young people with no right-wing attitudes.

Hypothesis B5: There are more young people who initially only passively consume right-wing extremist content than young people who actively post. The number of young people who network online lies somewhere in between.

According to research on some specific online activities among male adolescents (O’Malley et al., 2020), we put forth Hypothesis B6: In the field of right-wing extremism, the use of the Internet plays a bigger role for boys than for girls.

2.1 The Current Study

The quantitative youth study »JuPe (Perspectives of adolescents)« was conducted within the framework of the joint project »Radicalization within the digital age« and was funded by the German Federal Ministry of Education and Research. The youth study was developed within the framework of subproject II, carried out by the Criminological Research Institute of Lower Saxony, and focuses on the identification of potential risks and the exploration of vulnerable groups.

The aim of the youth study was to investigate socially relevant attitudes and behaviors in the context of radicalization. The survey was conducted in schools and provided information on a wide range of topics, including political views, religion, family, friends, and leisure activities. Psychological and sociological risk factors for radicalization, identified through previous research, were integrated (Goede et al., 2020). The questionnaire also included measures on extremist attitudes and behaviors.

2.2 Data and Participants

The study is based on cross-sectional data, which was gathered through an online school survey of ninth-grade students of all types of school forms except Special Educational Needs and vocational schools. In 11 of 16 federal German states, the authorities gave permission for the survey. The study used an oversampling of cities classified as having an extremism problem. We primarily focused on those cities that had been highlighted in reports by the Federal Office for the Protection of the Constitution, which is the domestic intelligence agency responsible for the detection of extremist attempts (German Federal Ministry of the Interior, Building and Community, 2022). Cities we selected had also been noted in police and media reports for their problems with Islamism or political extremism (Goede et al., 2020). Given that one of the original goals of the project was to analyze Islamist radicalization, we assumed a higher prevalence rate of extremist ideological beliefs in these cities and thus chose them in order to reach a sufficiently large number of relevant participants. Therefore, the sample is not representative of Germany as a whole.

The survey took place from January to December of 2018. All information was self-reported by the students. The questionnaire lasted around 90 minutes and was filled out online using a computer during school time. A trained test supervisor oversaw the standardized interview situation and a teacher was present. The parents were informed about the content of the study approximately one week before and had to provide consent for their child to participate in the study. Additionally, students had the right to not participate or to only answer some of the questions. Participation in the survey was voluntary and anonymous (Goede et al., 2020).

In total, 209 (19.1 %) of the contacted schools participated. Within the selected classes, 65.0 % of students participated in the survey. Following the data correction process, 6,715 from the originally 6,863 questioned students were included in the final sample. The participants were mostly 14 or 15 years old (M = 14.65; SD = 0.72). In terms of gender, 47.3 % were male, 52.5 % were female, and 0.2 % identified as neither male nor female or did not answer the question for other reasons. Of all participants, 43.4 % indicated a migration background, which means that the person or at least one parent was not born in Germany or does not have German citizenship.

2.3 Measurement

Reviews of the current literature show that there is no consensus on how to measure extremist attitudes (Beelmann, 2020; Goede et al., 2020; Lehmann & Jukschat, 2019; Schröder et al., 2020). We decided to measure extremist attitudes based on the definition of Beelmann (2020) but separately by phenomenon, which means that we use two dependent variables. The dependent variables are mean scales which contain items that are in accordance with the above definition. For this purpose, it was checked whether the respective items of the questionnaire were in line with the definition.

Since there are currently no validated scales that measure extremist Internet use, the authors have independently developed items for each examined extremist phenomenon. These original items are critically discussed in the conclusion.

2.4 Analytical Strategy

Descriptive results are presented first. Then, first for Islamism and then for right-wing extremism, the level of Internet activity is evaluated on the basis of mean comparisons, differentiated by extremist/non-extremist and gender. The dependent variables are mean scales with a range from 1 to 5. At point 3.00, these scales are dichotomized. Students with a value higher than 3.00 are considered to have extremist attitudes, since answer categories for the single items chosen higher than 3 (»4 – rather agree« and »5 – completely agree«) can be considered as agreement to the statements. The purpose of the dichotomization are mean comparisons of the groups. Due to the skewed distribution of the Internet items, the Mann-Whitney-U-Test is used.

To measure the intercorrelations of the different types of online activity and extremist attitudes, we conducted linear regression models with standardized beta coefficients and the two mean scales for extremist attitudes as dependent variables using IBM SPSS Statistics (Version 24). Therefore, not only extremist youth were included, but all participants, in order to take advantage of the entire variance of the data. For the area of Islamism, only Muslim youths could be considered in this study since the Internet items with Islamic content could only be answered by Muslims. For this, a filter was set in the questionnaire. All independent Internet items, as well as gender, age, socioeconomic status, and predicted school-leaving qualification, were entered as control variables.

Additionally, we calculated interaction effects to measure the interactions between gender and the examined main effects. Therefore, we z-transformed the relevant variables (Aiken & West, 1991). The interaction variables (Internet items X gender) were also entered.

3.1 Descriptive Results

Table 1 shows the descriptive results of extremist attitudes, control variables, and online activities. In this study, 14.8 % (n = 975) considered themselves to be Muslim. The analysis shows that 7.2 % of the Muslim students show Islamist attitudes (7.1 % among male students and 7.3 % among female students). In the following, we will refer to this group as Islamists. Of all Islamists in the sample, 47.5 % are male and 52.5 % female. Of all students, 6.4 % show right-wing attitudes (8.9 % among male students and 4.0 % among female students). This group is called the right-wing extremists and of those, 67.0 % are male and 33.0 % female. The distinction between extremists and non-extremists is carried out to subsequently compare the frequency of Internet use for political or religious motives between the extremist-minded youth and the total sample.

3.2 Analysis of Islamism

The purpose of Table 2 is to illustrate how frequently Muslim students use the Internet for Islamic or Islamist purposes, and which online activity is used most often. First, it can be shown that most of the Muslim students do not use the Internet for online activities with Islamic content, such as posting suras or quotes or meeting Muslims friends. For example, 71.4 % said they have never seen an Islamist violence video and 62.5 % have never met Muslim friends over the Internet. Of all Muslim students, 17.6 % »sometimes« or »often« meet Muslim friends via the Internet. Almost one fifth (19.5 %) »sometimes« or »often« posts Islamic content on their social media profile. Second, 8.6 % of the Muslim students »sometimes« or »often« watch videos with violent Islamist content. It can be deduced that active posting of Islamic content, followed by networking with other Muslims, is the most frequently used activity.

The purpose of Table 3 is to illustrate the mean values of each item representing online activities. We distinguished between adolescents with Islamist attitudes >3 on a fivepoint Likert scale and those with less or no approval to the statements. Mean values for the whole sample are reported as well.

The mean values indicate that online activities are not very prevalent in general. When the sample is split into extremists and non-extremists and by gender, the mean values range between 1.32 and 2.16 on a scale from 1 to 4. Differences between the groups are analyzed with Mann-Whitney-U-Tests. Extremists and non-extremists show significant differences for consumption and networking, but not for posting if gender is not differentiated (Table 3). The mean values of extremists are higher than those of non-extremists.

Differences between boys and girls are found among extremists for consumption and posting (Table 3). Girls have higher mean values among these items. Among non-extremists, only posting shows differences between boys and girls. Girls have a higher mean value here as well. When the whole sample is considered, it is again only posting that shows significant differences among genders, with girls having a higher mean value than boys.

¹ Islamist attitudes >3

The results of the linear regression models with Islamist attitudes as the dependent variable can be found in Table 4. Interactions between online activities and gender are included in Model 1 b. The results for Model 1 a (F(7,767)=21,761, p < .001) are mostly as expected, though socio-economic background shows no statistically significant correlation with Islamist attitudes. Gender (β = -.07*) is negatively related to Islamist attitudes, but the effect is very small. Also as predicted, school-leaving qualification (Abitur) (β = -.17***) is negatively related to Islamist attitudes. The lower the predicted qualification, the higher the Islamist attitudes. Consumption (β = .19***) and posting (β = .17***) are positively related to Islamist attitudes, which means that students who use the Internet for Islamic or violent Islamist content show more Islamist attitudes. Networking also shows a significant positive effect (β = .13**). The explanation of variance can be described as moderate according to Cohen (1988). Since Model 1 b (F(10,761)=15,506, p < .001) shows no significant effects for the interactions, it can be discarded. This means that there are no gender-related differences in terms of Internet activities and their effects.

3.3 Analysis of right-wing Extremism

Table 5 shows the frequencies of online activities with right-wing content of all surveyed adolescents. First, more than 75 % of the students do not use the Internet to consume right-wing content, whilst 7.6 % do this »sometimes« or »often«. Second, 5.8 % of the students »sometimes« or »often« meet people with right-wing attitudes via the Internet, and 3.9 % of the students »sometimes« or »often« post right-wing content on social media. From this it can be deduced that young people most often use the passive form of online activity, namely consuming right-wing content. Active posting of right-wing content is done only by very few young people.

The mean values indicate that online activities are not very prevalent in general. When the sample is split into extremists and non-extremists and by gender, the mean values range between 1.11 and 1.96 on a scale from 1 to 4 (Table 6).

Extremists and non-extremists show significant differences according to Mann-Whitney-U-Tests for consumption, networking, and posting if gender is not differentiated (Table 6). Extremists show higher mean values than non-extremists. Differences between boys and girls are found among extremists for consumption and networking (Table 6). In both cases, boys show higher mean values than girls. Among non-extremists, only networking shows differences between boys and girls. Boys show higher mean values here as well. When the whole sample is considered, all three areas of online activities show significant differences between genders. In all three areas, boys show higher mean values than do girls.

¹ Right-wing attitudes >3

The results of the linear regression models for the relationship between right-wing attitudes and online activities with right-wing content are shown in Table 7. Model 2 a shows the coefficients for the independent variables without interaction terms of online activities and gender (F(7,5389)=182,979, p < .001). Model 2 b considers the interaction terms and one such interaction is identified (F(10,5386)=132,214, p < .001). In Model 2 b, the predicted school-leaving qualification Abitur (β = -.21***) shows a negative effect on extremist attitudes. Students who are posting (β = .17***) right-wing content are more likely to show right-wing attitudes. The effect of networking on right-wing attitudes is not significant. Exploring the interaction between the three different Internet items with gender, one can see that gender decreases the effect of consumption on right-wing attitudes (β = -.20***). This means that the impact of consumption is lower on right-wing attitudes among females and higher among males. The general effect of gender is negative in this Model, which indicates higher extremist attitudes among males. The adjusted R² with .191 for Model 2 a and .196 for Model 2 b is, according to Cohen (1988), a moderate explanation of variance.

To visualize the significant interaction found in Model 2 b, the effects of consumption on right-wing attitudes were differentiated by gender (Figure 1). The graphical representation of the interaction effect clearly shows that, for boys, frequent consumption of right-wing content on the Internet leads them to display stronger right-wing extremist attitudes than would be the case for girls.

Figure 1

Effects of consumption among male and female students on right-wing attitudes.