Defining online aggression

As a multifaceted phenomenon, the concept of human aggression has been commonly associated with any behavior directed at another individual and carried out with the intention of causing harm to someone who does not wish to be harmed (Bandura, 1973; Berkowitz, 1993). Traditionally, the focus was on physical forms of aggression, as physical injury was potentially inflicted. With the increasing complexity of human relations and social interaction, non-physical forms of aggression (verbal, psychological, emotional, among others) resulting in various types of harm have been considered. Male levels of aggression tend to be higher than female ones, at every age and for all types of acts (Kruttschnitt, 2013; Salvatore and Markowitz, 2014).

Little’s taxonomy (Little, Henrich, Jones, and Hawley, 2003) presents four subtypes of aggressive behavior combining two axes. The first axis considers the forms of aggression: overt, involving direct confrontation with the victim; relational, damaging the victim’s social status or reputation and isolating her or him from social groups to which she or he belongs. The second axis considers its functions: proactive, based on the evaluation of consequences and anticipated goal of causing harm; reactive, essentially emotional and impulsive. Rieffe et al. (2016) suggest that children can distinguish their own motives for reactive and proactive forms of aggressive behavior.

The complexity of contemporary online life has raised new questions, since the faceless nature of digital interactions creates more personal ambivalence than do face-to-face encounters (Goldsmith and Brewer, 2015). The existence of an intention to cause harm to a specific target and of a potential target motivated to avoid such harm were traditionally key conditions for labeling any behavior as aggressive (Anderson and Bushman, 2002). As the perception of intent is essential to this process, the same aggressive act could be perceived in different ways, since it is dependent on individual judgment factors (Bandura, 1973). While time and space are becoming less synchronous, norms have also become more fluid and the strict boundaries between what is considered to be wrong and right are frequently questioned (Holt, Brewer, and Goldsmith, 2019).

Often, online aggression is subtle, and it may be harder to observe intention and all its outcomes for the potential victims (Amichai-Hamburger, 2013). The term “online aggression” as used in the EU Kids Online questionnaire designates hurtful interactions mediated through online technologies, whether or not repeated over time. This may encompass a wide range of harassing, threatening, embarrassing, unwanted and/or excluding behavior toward another individual, based on the subjective perspective of the respondents – in this case, children who perceive their online actions as aggressive.

A developmental approach to adolescent online aggression

A development-oriented approach to the study of aggression, particularly concerning its continuities and discontinuities across the lifespan (Eisner and Malti, 2015; Hartup, 2005; Piquero, Carriaga, Diamond, Kazemian, and Farrington, 2012), becomes relevant as this article considers the development of children and adolescents who present themselves as online aggressors and their interactions with others.

The concept of developmental tasks was introduced by Havighurst (1972[1948]) to designate socialization goals that must be achieved in particular biographical transitions. Expanding on Havighurst’s concept, Heymans (1994) defines developmental tasks as a period during which the individual has the opportunity to prove or to make plausible to a specific audience that she or he is capable of performing certain actions. An illustration is the social media environment, in which self-disclosures and self-presentations are continuously (re)constructed through what children and adolescents “explicitly provide, through what their friends share and as a product of how other people respond to them” (boyd, 2014, p. 49).

In spite of changes in the social and cultural contexts, some developmental tasks pointed out by Havighurst remain up-to-date regarding adolescence: adjusting to one’s developing sexuality, building a coherent identity, establishing intimate relations with peers and romantic partners, and achieving emotional independence from parents and other adults.

Both mass and social media act as contexts for those developmental tasks (Subrahmanyam and Smahel, 2011; Valkenburg and Piotrowski, 2017). While mass media-driven templates such as drama, reality shows, or celebrity news provide experiences through which young people learn how attention operates, social media play a crucial role in terms of social acceptance (boyd, 2014) and of co-creation of online contexts (Subrahmanyam and Smahel, 2011). The affordances of social media (the persistence of online expressions and content; the potential audience who can bear witness; the ease with which content can be shared; the ability to find content) create both opportunities and challenges for their networked publics – and when attention becomes a commodity, interpersonal conflicts emerge and battles over reputation, status, and popularity can be intentionally or accidentally hurtful to others (boyd, 2014).

For children and adolescents, the meanings of online problematic situations sometimes correspond to blurry categories. While aggressive communication encompasses “swear words”, “bad language”, “calling names”, and “cursing”, some stories of aggressive communication with friends are perceived as innocent jokes connected with the offline world (Smahel and Wright, 2014).

The victim–aggressor overlap

Often, victimization and aggression have been examined as dichotomized analytical categories (Eisner and Malti, 2015; Martinez-Ferrer, Moreno, and Musitu, 2018). However, since the mid-20^th century, a growing body of literature has shown that the same child may be engaged in both roles; therefore, labeling processes are critical (Farrington, 2020; Gaffney et al., 2016; Van Gelder, Averdijk, Eisner, and Ribeaud, 2015). Only some of the aggressors were/become victims of aggression, and not all victims end up engaging in aggressive conducts, but both share similar demographic and individual-level characteristics (Jennings, Piquero, and Reingle, 2012; Ttofi and Farrington, 2012). This explains how victims are more likely to be aggressors than non-victims and aggressors are more likely to be victims than non-aggressors (Lauritsen and Laub, 2007; Van Gelder et al., 2015).

Studies on children involved in different roles in online aggression (victim, perpetrator, or both), and the transitions between these roles, take account of the sociopsychological individual characteristics (Finkelhor, 2008; Van Gelder et al., 2015; Zych et al., 2020). Among those, low self-control is a major predictor of online aggression and of the victimization–aggression overlap (Agbaria, 2020; Bossler and Holt, 2010). Low self-control is more associated with short-sighted, impulsive, and insensitive decisions based on inaccurate judgements and misperceptions concerning the consequences of personal actions (Berg and Felson, 2016; Ousey, Wilcox, and Fisher, 2011). As neuroscience studies (Noble, 2017; Steinberg, 2017) reveal, adverse experiences in childhood impact the development of self-control, impulsiveness, and empathy, while the adolescent brain is highly susceptible to reward and peer influence.

A growing body of literature suggests that children who transit from the role of victim to the one of aggressor and vice-versa tend to show a more unstable developmental trajectory compared to those who remain in the same role (victim or aggressor) over time (Zych et al., 2020). Some authors suggest a negative relation between victimization and aggression (Ousey et al., 2011). However, victims are not a homogeneous group, and while some are passive subjects of aggression, others are not (Van Gelder et al., 2015). Furthermore, victimization affects different children differently (Finkelhor, 2008). Most empirical studies show that aggressors have a higher risk of victimization. The experience of aggressors having been victimized first is more common than the experience of aggressors transitioning from aggressive behavior to being victims. As the consequences of the risk of victimization during childhood affect children’s personality development, mental health, and academic performance, and have implications for the development of delinquent and antisocial behaviors, Finkelhor (2008) identifies mechanisms that affect the sequential processes of instigation, selection, and protection from aggression at the environmental and victim levels (Table 1).

Research on the victim–aggressor overlap among internet users has mainly focused on cyberbullying (Martinez et al., 2018), although not all aggressive behavior is bullying (boyd, 2014; Goldstein, 2015). To the classic definition of bullying (Olweus, Limber, and Mihalic, 1999) as involving psychological, physical, and/or social aggression, repetition, and imbalance in power, definitions of online bullying have added anonymity and public versus private action (Ybarra, boyd, Korchmaros, and Oppenheim, 2012). However, labeling interpersonal conflicts (such as individual acts of harassment, one-off fights, or reciprocal acts of relational aggression) as bullying “does not help teens navigate the complicated interpersonal dynamics and social challenges that they face”, says boyd (2014, p. 136), who appeals for attention to be given to teens’ language and cultural norms. Attention should also be paid to “the characteristics of bullies and victims, the relation between offline and online bullying, and consequences of such victimization” (Subrahmanyam and Smahel, 2011, p. 223).

Exposure factors for cyberbullying are the anonymity provided by the internet, the relational nature of the aggression, the high frequency of internet use, advanced internet skills, or the willingness to share personal and intimate information online (Dooley, Pyzalski, and Cross, 2009; Heirman and Walrave, 2008; Mesch and Talmud, 2010). Boys commit more cyberbullying than girls, while girls are more likely to be a cyber- as opposed to a face-to-face bully (Görzig and Ólafsson 2013). There is no significant gender difference in cybervictimization (Sorrentino, Baldry, Farrington, and Blaya, 2012). Age is a significant predictor of cyberbullying, and children holding a higher belief in their internet abilities are significantly more likely to bully on the internet and mobile phone than exclusively face-to-face (Görzig and Ólafsson 2013).

While cyberbullying has been explored fairly often, there is a relative lack of attention given to the nuances of the interaction between victims and perpetrators, and to the affordances of different platforms that facilitate this victimization (Seetharaman and Bhatt, 2019).

Children and adolescents’ “digital drift”

Recent criminological research targeting online offending and delinquent acts also offers a potential contribution to this analysis. Based on Matza’s (1964) drift theory, the concept of “digital drift” (Goldsmith and Brewer, 2015) suggests that everyday personal-computer and internet use may expand children’s interactions and opportunities for “drifting digitally” in and out of aggressive encounters, since identity flexibility is offered and pro-social influences seem less effective due to the relative absence of capable guardians.

Two key features of the “digital drift” are relevant to the discussion presented herein. First, the children’s self-reported aggressive behaviors are seen as a result of the intensification of online risks and opportunities in children’s lives, as they are increasingly engaged with internet platforms for multiple educational and recreational purposes. Often, children’s behaviors in physical and digital social worlds are intertwined and supported by transactional links with each other (Brewer et al., 2018; Subrahmanyam and Smahel, 2011). Furthermore, the new de-territorialized encounters and opportunities for online aggression cannot be dissociated from their relation to the nature of social control and pro-social influences in adolescents’ lives, nor from the latter’s previous experiences of victimization. They may not only be more exposed to, but also be more involved in, digital risk-taking actions and aggressive communication as victims, aggressors, or in the dual condition of victims-aggressors. Even if episodic for the vast majority of children and adolescents, aggression in digital media tends to become unpredictable and simultaneously often easier to carry out (Brewer et al., 2018; Goldsmith and Wall, 2019).

Second, affordances of online contexts may also favor online aggressive behavior by adolescents: the ease of use, a sense of competence or mastery, the speed of access, the absence of charge, an uninhibited behavior favored by anonymity, an overstimulation that favors a level of vertigo, a sense of unreality, the curiosity regarding the exploration of difference, and the asymmetry in the human-technology relation due to the lack of full individual control over the internet design (Goldsmith and Wall, 2019). However, what tends to separate aggressors from non-aggressors is the nature of the control placed upon each child’s life, whether at the internal level, by oneself (self-control), or from external sources at the micro and meso levels, such as family, peers, friends, school, and the community’s individuals and groups, or at the macrolevel, by the law, other institutional groups, and society.

Measures and identification of profiles

The definition of profiles considered variables related to internet access and uses, on the one hand, and engagement in online aggression and victimization, on the other (see Table 2).

The measures of internet access and uses comprise, firstly, the frequency of access to the internet by mobile phone or smartphone and by desktop computer, laptop, or notebook computer. Participants were asked to position themselves on a frequency scale of seven points (from “never” to “almost all the time”), which was then transformed into three points. Secondly, the time spent on the internet was computed based on the children’s estimation regarding weekdays and weekends. The average time was recoded into five points (from “up to one hour” to “seven hours or more”). Thirdly, a scale of digital skills derived from 11 statements was recoded into three categories: low, medium, and high.

The measures of online aggression and victimization take into account the frequency of aggression by means, namely in person, face-to-face, via mobile phone or internet, and in some other way (e. g., voice messages or text) as well as the frequency of having been treated in a hurtful or nasty way online in the previous year. In both cases, the original scale was recoded into three points (never, a few times, and at least every month). The differences in frequency allow us to distinguish between sporadic aggression/victimization and aggression experiences of a more repetitive and prolonged nature.

The results point to three patterns of aggressive online experiences organized in two dimensions (see Annex 1). The first dimension refers to the experiences of victimization/aggression, namely the means and frequency of aggression and the frequency of being subjected to aggression. The second dimension concerns the means and frequency of access to the internet, the time spent online, and a digital-skills scale.

In Figure 1, the categories of low and no aggression (face-to-face or other) and victimization are displayed on the right side of dimension 1; in dimension 2, the categories of an intense internet use and high digital skills (lower side) are opposite to a lower or lack of internet use and medium to low digital skills (upper side).

Figure 1:

Projection of children’s online aggression profiles.

The respondents were classified into three groups through a cluster analysis. For a start, we analyzed the descriptive measures of the variables that contributed to the definition of profiles by cluster (Table 3), in order to describe and name them. We moved on to explore their individual characteristics, namely age, gender, and psychological traits, particularly those related to self-control. Then, we analyzed online activities and exposure to online problematic situations. Finally, we considered the social mediation provided by family, peers, teachers, and other adults.

Emerging users, emerging aggressors

Children in cluster 1 are the least numerous and have the lowest ratios of access to devices, duration of use, and reported skills. Two-thirds reported having practiced online aggression sometimes in the previous year, and almost half had been victims of negative online situations. More than four in ten are 9 to 11 years old, immediately followed by those aged 12–14. Their gender ratio (slightly more boys than girls) is similar to cluster 3.

Emotionally, around one-third reported often experiencing unhappiness, lacking self-control, and being accused of anti-social behavior; one in five reported they often did exciting things even if those things were dangerous.

Although they present the lowest rates of online confidence and sense of online safety, a significant number was engaged in communicative practices involving self-exposure to people they may not have previously met face-to-face. In the previous month, four in ten had talked with people from other countries, and one in three had participated in online groups sharing similar interests or hobbies. Searching contacts had led around a quarter to actively look for new friends and add people to their contact lists they had never met face-to-face. One in five had created videos and uploaded them for sharing, which suggests a wish to attract others’ attention as well. Furthermore, one in three often found it easier to be themselves and talk about different things online rather than face-to-face. Meanwhile, some had been online victims in the previous year: Around one in five had lost money due to being cheated online, spent too much money on purchases and games, and experienced abuse of his or her identity.

Having reported slightly more bothering online situations than the other clusters, for children in cluster 1, parents and peers were the main social support – this is the single cluster that places parents first. In this regard, a third pointed to their siblings, while a quarter pointed to other adults they trusted; professionals whose job is to help children were more reported than in the other clusters. One in five children did not talk to anyone, similarly to other clusters and in line with the global results (Smahel et al., 2020).

Although parents are a key source of support in relation to bothering situations, and despite most children considering feeling safe at home and being helped by their families, more than half reported that their own voices were often ignored at home. The gap between children and parents is even wider on digital issues: More than half pointed out that parents neither encouraged them to explore and learn things on the internet nor talked to them about what they did online. One third reported parental use of technological devices to monitor their behavior, which is the highest rate out of the three clusters. Nevertheless, almost four in ten children often acted as digital helpers for their parents.

High users, high aggressors

Cluster 2 refers to children who reported the most aggressive practices and the most intense victimization, online and offline, spending the most time on the internet, using the computer the most and estimating the highest digital skills. Mainly aged 12+, seven in ten are boys. They present the highest emotionally and socially problematic behaviors: More than half reported doing exciting things even if they were dangerous – slightly more than those who did dangerous things for fun –, losing self-control, and experiencing unhappiness. Around one in three was also accused by others of lying and cheating.

Online, they are by far the most engaged in almost all communication and participatory practices, in which they expose themselves and take risks related to their privacy and safety. Six in ten looked for new friends or contacts at least every month and participated in online groups where people share their interests or hobbies. These practices were followed by talking with people from other countries and adding people they had never met face-to-face to their contacts. Around a third were regularly engaged in content creation for sharing and participated in discussions on political and social issues. In the previous year, almost four in five had seen sexual images, which is the highest value out of the three clusters.

Around seven in ten often/always felt safe online and were convinced they knew what to do if someone acted in a way they did not like. Furthermore, almost half found it easier to be themselves online rather than face-to-face and considered that other people were kind and helpful on the internet. They strongly trusted the online environment and their own skills for exploring their identity and anonymity, although a quarter had experienced violation of their data and situations of identity abuse. They are close to cluster 1 in regard to having lost money and having been cheated in the previous year.

When they experienced bothering or upsetting situations, peers doubled parents as the main source of support. Teachers, professionals whose job is to help children, and other adults they trusted were less reported than in cluster 1.

Regarding communication, support, and safety at home, this cluster occupies an in-between position. More than three in five felt safe at home and recognized family support as fairly/very true, a rate slightly higher than the rate of those who considered they were listened to when they spoke. Communication is also an issue, since more than half reported that their parents hardly ever talked with them about what they did online (a situation that was frequent for only a fifth of the children). Being encouraged by parents to explore the internet and learn things never occurred for almost half of the children, while it was true for around a quarter. Almost half were often the digital helpers at home. Parental controls for blocking, filtering, or tracking contents have been reported by around a fifth of the children.

High users, occasional aggressors

Children in cluster 3, the largest one, reported having engaged in a few aggressive practices in the previous year, online and face-to-face. Half of the individuals had also occasionally experienced victimization. More than half are frequent internet users, accessing it for 3–6 hours a day and self-reporting high-level digital skills.

Nine in ten are aged 12–16, and boys are slightly more numerous than girls, as in cluster 1. Around a third reported frequently losing self-control and doing exciting things even if those things were dangerous, while a quarter pointed out frequent feelings of unhappiness and risk-taking behaviors for fun. Antisocial behaviors (robbery, lying, and cheating) were far less reported than in the previous clusters.

Children in cluster 3 also reported frequent activities involving communication, engagement, and content creation at least once a month, and almost all rates are slightly similar to those of cluster 1. In line with their developmental tasks, the largest difference is related to participation in online groups where people share their interests and hobbies, which was reported by almost half of the children; almost three in four reported they had seen sexual images in the previous year.

As in cluster 2, around two-thirds of the children felt safe online, almost half considered that online people were kind and helpful and – above all – three in four were confident that they knew what to do if someone acted online in a way they did not like. Cluster 3 presents the lowest rates of experiencing data misuse and problematic situations related to money.

Around a third found it easier to be themselves online and to talk online about different things that they silenced face-to-face. These rates are almost twice the rate of those who often talked about personal things which they would not talk about face-to-face (18 %, the lowest rate). In line with a more cautious behavior regarding privacy and identity management, cluster 3 presents the lowest values regarding sending personal information or pretending to be a different kind of person.

Three in five children reported having experienced bothering or upsetting situations in the previous year. When this happened, peers (62 %) more than doubled parents (30 %) as the main source for advice. One in ten reported having talked with adults they trusted, placing them above teachers and professionals whose job is to help children.

Cluster 3 presents the highest level of perceptions concerning their family’s attention, support, and safety. However, almost half reported that their parents neither encouraged them to explore and learn things on the internet nor talked with them about what they did online. Presenting the lowest rates regarding the use of parental controls for blocking, filtering, or tracking content, these adolescents are undoubtedly seen as the family’s digital helpers.

Cross-view of the clusters

Low self-control emerges as the most consistent correlate of online aggression, aggravated by the ineffectiveness of external controls in children’s lives, both at the informal and formal levels.

All clusters share high levels of emotional deprivation, which is visible in the children’s low levels of well-being and self-control as well as (possible) social isolation translated into their search for exciting contacts, attention, visibility, and social acceptance. Children also present an exacerbated sense of online safety and trust in their own skills for dealing with problematic situations despite the victimization experiences. Furthermore, they face a relative social control deficit and poor/inadequate digital supervision from their parents; teachers and professionals whose work is to help children are almost absent. This may be decisive for the way in which they anticipate and perceive their behaviors to be seen by others, which facilitates the non-internalization of conventional internal controls.

Although the clusters share a male predominance, the gender gap is not as wide as reported in most studies on offline aggression (Kruttschnitt, 2013; Salvatore and Markowitz, 2014).

Mainly composed of adolescents fully engaged in their developmental tasks, clusters 2 and 3 clearly differ from each other. A frequent engagement in aggressive acts and high levels of problematic emotional characteristics, transgressive practices, recognition of others’ criticism, and online performances related to communication and participation portrays the individuals in cluster 2 as “omnipotent masters of the digital”, who consider they can easily do whatever they want. In cluster 3 (the largest one), occasional aggression seems to be related to relational online peer victimization. Presenting lower vulnerable psychological traits and deviant practices, these individuals are also the most competent regarding their own online protection and privacy.

Individuals from cluster 1 may in the future fall into one of these clusters. The risk of falling into the frequent aggressive practices evidenced by cluster 2 is relatively high, due to their psychological traits and signs of risky behavior as well as to the lack of parental supervision and communication, which are replaced by technological devices.

The analysis confirms that age is a stronger predictor than gender among children who reported online aggression. Not only do the older ones present high levels of risk-taking behaviors, but they also report low levels of adult supervision and communication. The relation between internet uses and engagement in online aggression is not straightforward, suggesting the prevalence of personal and social traits over online access and use.