Keywords

1 Introduction

Social media is defined as “computer-mediated communication channels that allow users to engage in social interaction with broad and narrow audiences in real time or asynchronously” [1]. They are an integral part of our everyday lives offering diverse opportunities for communication and interaction way beyond what was possible only a few years ago. 98% of young users between 15 and 24 years use the Internet daily [2]. Most of them (96%) use social media, where they share experiences, opinions, and news. According to a representative study with 1,316 U.S. teenagers aged between 13 and 17 shows that the most frequently used social media are TikTok, Instagram, and Snapchat [3]. These networks are used even more often than a few years ago. While in 2015 52% of teenagers used Instagram, 62% of teenagers are using Instagram in 2022. The usage of Snapchat increased from 41% to 59%. TikTok, however, was not available in 2015 which is why we cannot report an increase from that time. Still, it is the most prominent “classic” social network nowadays, with 67% of teenagers using it. The use of Facebook significantly decreased from 71% in 2015 to 32% in 2022. The only social media platform that is used more than TikTok is YouTube (95%). However, the usage patterns on YouTube differ slightly from those of TikTok, Instagram, and Snapchat, where direct communication between users is more important than on YouTube.

The ubiquity and usage intensity of different social media platforms causes that even children and adolescents are profusely confronted with very heterogeneous, and sometimes harmful content. Content, network and algorithms are factors of social media that may also represent threats for (young) users. Such threats can be digital addiction, discrimination, hate speech, misinformation, and polarization, manipulative influences of algorithms as well as body stereotyping or cyberhate and cyberbullying [4].

Consequences of engaging with such toxic content and dynamics can range from negative affect, reduced well-being, and emotional problems to severe depressive symptoms and suicidal thoughts [5,6,7,8]. Furthermore, victims of cyberbullying or aggression might become a bully themselves as a kind of exit strategy [9]. Notably, victims as well as bullies suffer from the bullying situation [10], and both are shunned from their classmates [11]. It becomes clear that bullying (as an example for toxic dynamics) affects a whole social environment instead of only the victim and the perpetrator, for instance the class in school. This can harm the cohesion and healthy group dynamic in class and beyond.

Young users can be overwhelmed by the flow of content and do not know how to protect themselves from harm or how to help other users (see [12]). Consequently, the educators, caregivers, and social organizations (among others) are challenged to help students mitigate risks by conveying critical skills and social media literacy. Notably, we do not want to ban the use of social media at all. This would not be a realistic approach since social media are ubiquitous and users also say that they would have a problem with quitting social media. More precisely, 54% say that it would be “very hard” for them to stop using social media, 18% say it would be “hard”, and another 35% say it would be “somewhat hard” [3].

A more promising and sustainable approach is to raise awareness of this issue and to increase social media literacy. When teaching social media literacy, which includes “decoding, evaluating, and creating” social media content [13], it is important that users’ are not only provided with information or pre-defined solutions to a problem, but that they learn how to implement strategies and adapt their knowledge to different contexts and situations [14]. Thus, it is not only about factual or declarative competence, but also about procedural competence [14,15,16]. This approach goes hand in hand with increasing critical thinking abilities of users. Further relevant aspects for developing social media literacy cover self-protection skills, emotional intelligence, empathy, and should be driven by intrinsic motivation. In all the mentioned endeavors it is crucial to not violate users’ autonomy. However, developing social media literacy and self-protection skills needs time and effort. As long as students do not have sufficient skills for acting fully self-determined, they need support (without coercion). This support can be provided by interweaving technical tools with guidance by educators, ensuring students’ autonomy, increasing their competence, and providing an experience of social relatedness (see self-determination theory, [17]).

In line with this, teachers and educators have an increasing responsibility in providing and implementing learning activities that stimulate reflection on the mechanisms behind the use of social media [18, 19]. However, most of them are not adequately prepared to face these challenges, and, consequently, there is an increasing need in providing educators with new methodologies and tools specifically designed for these purposes [20]. Teacher education would benefit from including critical social media literacy since social media spaces are not neutral and students need strategies and tools to leverage the opportunities emerging in these spaces [19, 21]. In [22], for instance, the ecological system theory has been leveraged to affect the school microsystem and assist students through specific training activities and innovative tools in understanding the mechanisms and algorithms of social media platforms with respect to toxic content propagation, content recommendation and personalization. In the following, we will a) report findings from psychological literature demonstrating the severity of cyber aggression and its impact on users’ well-being, and b) refer more detailed to the role of educators for increasing students’ social media awareness and self-protection skills, then we c) introduce the COURAGE projectFootnote 1, in which a social media companion safeguarding and educating students is proposed, aimed at supporting educators in promoting innovative educational interventions to increase students’ awareness on social media mechanisms and the consequent threats that in these environments proliferate. The impact of the COURAGE project actions in educational practices is finally presented.

2 The Severity of Cyber Aggression for Social Media Users

The systematic review in [4] points out, there is no comprehensive and commonly adopted definition of cyberbullying and/or cyberhate. Thus, we refer to any form of cyber aggression on social media (e.g., cyberbullying, hate speech, fake news) when talking about toxic content on social media in our project. However, we argue that it is important to be very clear about the definition of cyber aggressive behavior when analyzing it in an empirical study in order to be confident about the causes and effects.

Cyberbullying interactions on social media can be particularly damaging because they are permanent, easy to spread, and can appear at every time of the day, making monitoring and moderation a particularly expensive task also from the psychological side [23]. Moreover, teenagers may spend a substantial amount of time on social media, which can strongly affect them not only as victims or bullies but also induce a change of attitude as a consequence of observing such events.

Prior studies measuring the impact of cyberbullying, for instance, showed that bullying on social media can be related to severe problems such as lower self-esteem, perceived loneliness, psychosomatic or depressive symptoms [24,25,26]. The influence of both, cyberbullying, and bullying in schools disentangled from social media has been investigated in [27]. Both forms of bullying were revealed to be severe risk factors for post-traumatic stress symptoms of adolescents between 11 and 18 years (μ = 15.6, σ = 1.89). Notably, many cyberbullying scenarios have their origin in disputes in the classroom, or, the other way around, bullying in school contexts can be expanded to social media and reach an even larger audience. Therefore, it is even more important to take effective action against cyberbullying and cyberhate that could contrast the extended exposition that unprepared teenagers can have online. Schools are the perfect place to raise awareness about the three shared characteristics of cyberbullying and bullying outside social media: aggression, power imbalance, repeated behavior [28], and also about social media features which are actually able to reinforce cyber aggression (e.g., perceived anonymity, increased number of bystanders, persistence of the bullying content, and continuous access to the platform). However, providing sustainable additional support outside of the classroom is also likely to be necessary especially for the most fragile users.

Authors in [29] investigated the effects of hate speech by means of six million comments on Reddit provided in 174 online college communities and revealed that being confronted with hate is related to increased expression of stress. However, their results also reveal that the effects depend on individual characteristics of the users, for example, their psychological endurance (i.e. resilience). Resilience comprises adversity and adaptation, and can be more or less stable [30]. However, it definitely is an important factor for dampening the effects of cyber aggression. Authors in [31] also investigated the role of resilience in the context of hate speech and found that resilient hate speech victims report lower depressive symptoms than victims with lower expression of resilience.

Besides the severe influence on adolescents’ well-being and psychological health, toxic dynamics on social media are also related to poorer academic performances. Kowalski and Limber in [28], for instance, showed significant correlations between poor grades, absence from school, quantity of leaving school on the one side, and being involved in cyberbullying (as perpetrator or victim) on the other, for students between 11 and 19 years (μ = 15.16, σ = 1.76). Here, we need to take into consideration that authors in [28] report correlations instead of causal effects. Thus, it is actually not clear whether bullying leads to poor grades, or poor grades have an impact on bullying. Nevertheless, acting against bullying – no matter on which side of the causal chain – can help to overcome a toxic relationship.

When looking at cyber aggression on social media, there is not only one perpetrator who is the reason for the spread of toxic content, the bystanders play a significant role as well. Users who are observing cyberbullying scenarios, hate speech, discrimination or other harmful dynamics on social media without taking action might contribute to the spread of such content, albeit involuntarily and without malicious intent. Studies reveal that the intention to intervene in problematic scenarios is related to users’ empathy and the perceived similarity to the victim [32]. Thus, if the bystander is able to mentally put himself in the other person’s place, there is an increased possibility that the victim receives help from the bystander. Therefore, empathy and emotional intelligence are severe factors for acting against toxic dynamics on social media. According to a large meta-analysis covering 146 different definitions of empathy by Guthridge and Giummarra [33], empathy can be defined as “the ability to experience affective and cognitive states of another person, while maintaining a distinct self, in order to understand the other.” Emotional intelligence resembles, amongst others, the ability to understand the emotions of other people, to use emotions to facilitate decision-making, and to manage emotions [34]. Interestingly, emotional intelligence itself was already shown to be positively related to mental health, well-being, and academic performances [34, 35]. Thus, it is not only important for preventing cybermobbing, but for educational contexts in general. An important aspect is also the ability to effectively react for a victim of cyberbullying and intervene in their support, as ineffective reactions may worsen the situation. Indeed, different aspects of emotional intelligence, (i.e. intrapersonal and interpersonal emotional competence) might play different roles and is important to provide specific educational support to develop each of them [36].

3 The Role of Educators to Increase Students’ Social Media Awareness

The role of teachers in promoting social media, digital literacy and countering disinformation has been recently highlighted by the EU commission in the final report of the “Commission expert group on tackling disinformation and promoting digital literacy through education and training” [37]. In this report, initial teacher education (ITE) and continuing professional development (CPD) have been identified as fundamental elements to foster teachers' intervention. Also, educational online applications are crucial to support teachers in developing digital and media literacy competences. Innovative educational approaches and tools are needed to promote educational interventions aimed at providing students with the needed tools to counteract and face social media threats. For instance, in [38] researchers showed that students revealed an increasing ability to judge the credibility of digital content when they participated in lessons in which, by experiencing lateral reading, they learnt how to evaluate online content and identify fact news. This confirms the central role of educational interventions in changing the way students approach social media threats and improving their well-being in social media environments. Educational interventions specifically focused on addressing these challenges have to be designed and also teachers need to be empowered to include these interventions within their educational strategies.

The need to ‘[empower] teachers so that they are able to take an active stand against all forms of discrimination and racism, to educate children and young people in digital literacy’ is a concept that has been proposed in March 2015 in the framework of the Paris Declaration of EU Education Ministers. The conclusions of the EU Council in May 2016 on the role of education concern how to provide learners with the competences and values required ‘to access, interpret, produce and use information and other media content, notably in the context of the internet and social media, in a safe and responsible manner’. In September 2020, the Digital Education Action Plan (2021–2027) of the EU Commission stated that ‘it is important to educate people at all ages about the impact of digital technology on well-being and the way technology systems work.’ In the meantime, empirical investigations highlighted the need to empower teachers and educators so that they can support their students, as well (e.g., [18, 19, 21]).

The report by the EU Commission (2022) highlighted the areas of digital literacy which should be addressed more in education and training, specifically the need for further training of educators to acquire the competences needed to engage their students in creative ways, the need to develop digital tools and environments not only for educational purposes, but also to promote individual and social well-being.

In this direction the COURAGE project provides solutions which can be introduced and implemented by educators in school contexts, until students are able to use the tools on their own, or – in the long term – until they have sufficient social media awareness, social media literacy, and self-protection skills. Importantly, we provide information, user friendly interfaces, and workshops that are helpful for both students and teachers.

4 The Courage Project

In the framework of the COURAGE project, the multinational and multidisciplinary team provides educators with new tools and learning methodologies that can be adopted within higher education learning paths to train educators in supporting students in facing the social media threats. The multidisciplinary nature of the project team was leveraged to face the problem from different perspectives. In particular, we propose novel approaches exploiting most recent advances in the fields of artificial intelligence and in the educational domain paired with social and media psychological insights. The basis for the tools and training-scenarios provided by the COURAGE project is a complex technical development resulting in the conceptual architecture described in [39]. Building on that, and on research questions and hypotheses derived from prior literature, various scenarios have been created and empirically evaluated by the project team. In the following, we will provide a compressed overview of the technical development (see also [39]), also referring to the role of AI, and then refer to some of our empirical investigations which are explained in more detail in [40].

4.1 AI, Media Literacy, and Societal Changes in the COURAGE Project

Artificial intelligence (AI) proposes new opportunities in educational support systems, but poses challenges (e.g., privacy threats, manipulation, threat to autonomy) and promising outcomes (e.g., suited user support) at the same time. AI can be viewed from different perspectives – for instance how it provides content to social media users in their social networks (by means of algorithms for microtargeting or recommender systems), how it provides information to users through a smart speaker device, or how it might help to improve the performance of technical tools for supporting users regarding a non-harmful social media use.

The influence of AI technologies in several aspects of media content production has led to the need of promoting media literacy education interventions aimed at unveiling the mechanisms at the basis of algorithm-driven media. To this aim, educators should elaborate specific learning paths connecting media literacy education and computing education [41].

In fact, if from one side the issues related to social media have been identified, from the other side there is a lack of appropriate educational interventions able to face the social media challenges and support students in developing the needed media literacy skills. In the framework of COURAGE specific tools and methodologies have been designed to this purpose, with the aim to provide educators with a comprehensive toolbox combining both technological support and also pedagogical strategies. The aim is to better address the issue of increasing awareness of students on social media threats. In this perspective, the approach proposed in COURAGE brings the educators towards the challenges that media literacy poses in the machine learning age. As stated by Valtonen et al. in [41], topics such as dynamic content generation, recommenders and optimization, content filtering, attention engineering, reinforcement learning are directly related to educational content aimed at developing skills and knowledge about how artificial intelligence applications operate behind social media platforms.

The research on artificial intelligence in higher education needs to provide more insights on how to progress on the pedagogical models and how effectively AI can support higher educational teachers to make teaching and learning more effective [42]. The categorization of AI applications proposed by Zawacki-Richter in [42] identified four categories: application for profiling and prediction, application to support assessment and evaluation, adaptive systems and personalisation, and intelligent tutoring systems. The COURAGE conceptual framework provides different contributions in these four areas with specific modules, in which the interconnection between the theoretical pedagogical strategies and the educational application of AI approaches plays a relevant role [43].

In our project, we consider AI as a chance to provide adequate support for users by focusing on the advantages of new technologies in a digitalized society. However, the potential threats of using AI as well as risks associated with AI used to manipulate users in social media environments shape our ethical and moral understanding regarding responsibilities that come along with the consideration of AI-based systems.

As AI algorithms of social media platforms do not directly counter threats for teenagers but may actually worsen and amplify them, in the COURAGE project we propose the adoption of a virtual learning companion (VLC) powered by an AI system that counters these threats and the platform specific algorithmic hazards [43]. The VLC AI system developed comprises adaptive detectors of content and network threats, user models to support personalized interventions as well as content and educational activity recommendations.

The VLC also supports educators in empowering teenagers regarding the threats in social media by developing social media competences and self-protection skills. A particular emphasis was put on the role that experts and educators play in the management of the VLC by driving and moderating the use of it in classroom-based educational activities as well as providing learning strategies, objectives and activities implemented in the VLC [43,44,45]. In the next section we present how the VLC has been piloted in real scenarios through different educational interventions.

4.2 Transfer of the Virtual Learning Companion into Practice

The current version of the COURAGE eco-system has been experimented in several studies with teachers and students of secondary schools. Overall we can rely on workshops, discussions with around 500 experts and students, as well as empirical studies, in Italy, Spain, and Germany.

The aims of these investigations were and still are to:

  1. a)

    analyze the effects of training activities about social media mechanisms (algorithms and social influence),

  2. b)

    support understanding and dealing with different types of threats (misinformation, filter bubbles, hate speech, ethnic and body stereotypes),

  3. c)

    encourage a constructive use of social media,

  4. d)

    assess procedures aimed at analyzing psychological characteristics of adolescents in correlation with social media use such as mental well-being, emotional intelligence, critical thinking, life satisfaction and resilience,

  5. e)

    implement scenarios of educational interventions through narrative scripts, minigames, and educational content design.

Overall, the technical environments and empirical studies demonstrate the feasibility and encouraging positive impact of the approach. In the following we will report some of the empirical investigations that have been conducted. For instance, gamification (i.e. “the use of game design elements in non-game contexts” [46]) strategies based on the concept of interactive narrative and counter-narrative (i.e. scripts that challenge previously made assumptions) scripts, have been implemented to support learning scenarios. In our studies, students were immersed in social media stories that exposed them to counter-narratives and conversations about counter-narratives, about biases, discrimination, or attitudes and behaviors in what (and how) is spread online [47]. These components direct content recommendations and educational interventions, such as triggering specific validation tests that improve the accuracy of user state estimation as well as progressing in narrative scripts, which are gamified educational activities based on the concept of interactive counter-narratives. Scripts and material were evaluated positively and were related to high levels of intrinsic motivation (μ = 4.41, σ = 1.25) and low levels of perceived pressure/tension (μ = 2.52, σ = 1.21), both on a 5-point Likert Scale. Against the background of self-determination theory [17] and AI guidelines stating that it is important and necessary to guarantee users’ autonomy, this is a satisfying result. This also resembles an important aim of the COURAGE project, namely, providing support without using paternalistic approaches, allowing for the development and enhancement of users’ interdependent self-protection skills. Not using paternalistic approaches means respecting users’ autonomy and self-determination through leaving the final decision and behavior to the user. Our approach shall guide the user towards the right (i.e. less harmful) direction, but not force her to perform or not perform a certain activity. A specific game based educational activity allowed students to directly experience and grow aware of the effects that phenomena like echo chambers and filter bubbles can have on their perception of the world and their decisions. This is particularly important as these phenomena are usually almost imperceptible for the users [48]. Further empirical works are reported by [40].

The efficiency of more advanced, multi-step user specific, algorithmic interventions, e.g., content and connection recommendations, to a) counter network and algorithmic threats, b) support the achievement of educational targets, and c) improve the collective well-being of the user community is currently supported by simulations based on collected data and validated user models [43].

In particular, we studied the relationship between users’ satisfaction, content diversity exposure and collective well-being metric through the simulation of three different recommender systems using three different strategies to: maximize opinion diversity, overlapping third order neighborhood and random recommendations. Satisfaction is assumed as a proxy for the sustainability of the social media platform. Content diversity exposure could play an important role in countering the effects of filter bubbles [49], echo chambers [50], and ultimately society polarization. The recommender that maximizes the diversity of opinion between the pairs of users to connect showed a slower start but achieved higher exposition to more diverse content and a similar level of satisfaction to the other two recommender systems.

5 Impact of the COURAGE project

The COURAGE project highlights the relevance and urgency of raising users’ awareness of social media threats, increasing their social media literacy and self-protection skills by means of both technical support and assistance to educators who can make use of tools developed within the project. Figure 1 provides a summary of the most relevant aspects of the COURAGE project, highlighting the role and application areas of educators. It shows that the interaction with the VLC (within the social media environment) is the starting point of the process aiming at developing and establishing self-protection skills of students. In a first step, the VLC helps students to develop and train a) relevant personal skills like empathy or emotional intelligence, b) social media / digital skills like using reverse image search for identifying the origin of a manipulated picture or considering specific (privacy) settings within a social network, c) knowledge regarding toxic content such as reasons and quantity of discrimination on social media or the severity of promoted “perfect” body images. These skills then should increase social media awareness and social media literacy and finally, students’ ability to autonomously protect themselves against the threats of social media. Educators should support the students in this process by introducing the VLC and related tools and accompanying them until they have sufficient skills to either use the VLC autonomously or until they are able to protect themselves, even without the VLC.

Fig. 1.
figure 1

Conceptual overview of the aims of the COURAGE project

Full size image

In addition, the project emphasizes the responsibility of teachers who would benefit from technical support when educating their students with regard to social media literacy. Besides having social media literacy themselves, educators also need to know how to transmit factual and procedural knowledge regarding a safe use of social media. Transmitting this knowledge to students is a responsible and far-reaching task, because it will affect users’ everyday as well as future life. Therefore, it is important to offer workshops and sophisticated accompanying material for educators who are confronted with this important duty. In fact, the success of transmitting social media literacy in educational contexts strongly depends on the educators’ skills and knowledge [51]. Since schools nowadays often are working with digital devices, training social media literacy should be a fixed component of curricula. Here, students and teachers could start working with the VLC and narrative scripts and continue using it in their everyday lives outside of school. COURAGE focuses exactly on these interwoven aspects: addressing young users and educators, which is an advantage in comparison towards other approaches focusing on students or educators. And instead of simply providing tools and materials, the project teams are going to schools in order to conduct workshops and receive qualitative feedback from students and teachers.

Current developments such as the increased use of social media due to the worldwide COVID-19 pandemic [52] reveal an even higher demand for action in this realm. Studies show that students are engaging with social media more often and more intensely since the outbreak of the pandemic (e.g., [53]). At the same time, parents, educators and caregivers had and still have plenty of other responsibilities so that children were and still are often alone with their social media, aggravating the threat of toxic content.

Furthermore, with our project, we are facing huge challenges in terms of normative and ethical questions. This has several reasons.

First, we are addressing young users who are in a sensitive phase of their lives and whose identities are still developing. Following the psychosocial theory by Erikson [54], adolescents deal with identity formation and identity confusion. This includes exploring and perceiving the social surrounding, learning from it, and developing attitudes and patterns. Depending on the social surroundings, role models and references can foster either a social, critical and empathic direction of development, or a toxic development. Since social media are an integral part of adolescents’ social life, there is a huge range of role models, examples and circumstances which might be an anchor point for them in their development. Here, we see the need for supporting users when facing toxic content and dynamics in order to provide more space for positive and meaningful experiences on social media which might be beneficial, or at least less harmful than discriminating, offending or manipulated content.

Second, we are working with AI, which brings several ethical challenges itself. The use of AI algorithms is connected to new types of ethical issues, when the algorithms are trained on biased data they will produce, in turn, biased output. Therefore, they cannot be considered impartial and their effects on discrimination, prejudice and stereotypes are amplified, and this is even worst when it comes to AI algorithms that regulate social media mechanisms, since these algorithms influence the ecosystems in which young adolescents grew up, thus often providing the lens through which adolescents shape their understanding on new concepts.

Third, when it comes to content on social media, it is really hard to decide what is acceptable and what is not. Who determines the border between good and bad? For some people, a posting might be severe and offending, for others not. Thus, it is not easy to overarchingly categorize content depending on its severity and harmfulness. Consequently, the indicators initiating assistance by the VLC need to be chosen very carefully. Besides setting the baseline for when to intervene and support, working with AI systems needs to consider further aspects like privacy protection, fairness, transparency, or common good [55], to name just a few.

6 Conclusion

We have already achieved important goals, since first qualitative and quantitative evaluations show that the VLC and the different VLC scenarios can support students as well as educators, by providing information, increasing skills and motivation, and training social media literacy as well as raising awareness of users. Users’ experiences in schools or social environments can affect their experiences on social media and the other way around. Thus, we are dealing with an overarching challenge for students’ everyday life and their future. We want to “make social media a good place”, to improve the overall experience of social media and help students in developing self-confidence in digital spaces and beyond.