Empirically Investigating Virtual Learning Companions to Enhance Social Media Literacy

Social media are “computer-mediated communication channels that allow users to engage in social interaction with broad and narrow audiences in real time or asynchronously” [4, p. 472]. The most common social network sites such as Facebook, Instagram, Twitter, and TikTok provide these elements and thereby allow for bonding with others, disclosing the self, and getting social support [14, 31, 34]. However, aside from these positive aspects and experiences, the use of social media can bring negative consequences, as well. This is alarming as a large percentage of young users aged between 15 and 24 years use social networking sites [16]. Leaving them exposed to potentially harmful content in their everyday lives, even during school hours.

Harmful content can be related to body image, identity, or any other aspects which are important to young users, and can occur by means of toxic dynamics like cyberbullying, hate speech, fake news, or discrimination [32]. Consequences of being confronted and engaging with such negative dynamics can be related to somatic or depressive symptoms [21, 58], emotional problems, poorer academic performances [54], or suicidal thoughts [59]. In addition to that, bullied individuals or those confronted with toxic content could evolve as bullies themselves, because they do not know how to protect themselves or get out of the situation otherwise. From a more technical perspective, algorithms as well as targeting processes can also be a threat to users’ autonomy (e.g., [30]).

Although social media is not bad per se, and there are many other predictors for low well-being or depressive symptoms [12], these threats call for action in researchers to formulate innovative ways and develop educational tools to increase awareness in young users with respect to the panoply of harmful phenomena of social media [42]. In the long term the aim would be to increase users’ self-protection skills, so that they can act completely autonomously and self-secure on social media.

Scholars have already investigated the role of social media literacy (SML) in the context of the secure use of social media [8, 18, 34]. Following Livingstone (2014, p. 286), SML includes and refers to “the tasks of decoding, evaluating and creating communication in relation to media qua representation (text, image, platform, device, etc.) and qua social interaction (relationships, networks, privacy, anonymity, etc.), since these are integrated in the very nature and use of social networking sites”.

Interventions focusing on SML have so far demonstrated positive effects towards skills acquisition [20, 41], leading the way to more technologically advanced approaches. One such approach is the implementation of web-browser plugins that act as virtual learning companions transferring information to users during content browsing [33]. Virtual learning companions usually take the form of chat-bots ready to support a user when in need without being intrusive. They can be deployed to accompany users through educational resources and provide further knowledge when requested.

We argue that SML supported by virtual learning companions can help to educate users to not only understand harmful content, but also the effects of the messages that one creates and distributes. This can serve as the key for creating and enhancing self-protection skills and guaranteeing autonomous and self-secure use of social media. In the COURAGE project we aim at inducing SML and several self-protection skills (including e.g., raising awareness, emotional intelligence, empathy) within the environment of social media supported by a virtual learning companion [44, 51]. This will provide users directly with support when needed, avoiding adoption problems. In this paper we report seven approaches towards SML skills acquisition powered by virtual learning companions, developed as part of the COURAGE project. A discussion is then made towards their potential benefits as potential educational tools.

SML is a form of literacy that focuses on training skills for the growing digital environment. A “correct” (i.e., non-harmful) usage of social media is crucial to exploit all the advantages offered by the new media, especially for teenagers who need to be supported in order to cope with the different threats and drawbacks omnipresent in it. SML interventions have demonstrated to decrease social media addiction [25], and to be beneficial towards the acquisition of skills related to digital competences [41] and fake news discernment [20] amongst others. They can be integrated in school curriculum as specific subjects or integrated within each curricular discipline. Such interventions can guide students in analyzing social media messages by reflecting on the sources, the language used, and the reliability of images promoting and reinforcing other literacies (i.e., health, arts, citizenship) [7].

Current studies have developed different forms to deliver SML interventions. For instance, an initiative led by [55] designed a collection of activities for classroom instruction to combat misinformation in social media. On their part, [47] aimed to raise awareness of body image concerns, by designing an intervention manual which was beneficial in the format of six or three interactive lessons [13, 41]. More recent interventions, such as the one conducted by [6] during the COVID pandemic found that online asynchronous assignments can help to improve fact-checking through a lateral reading course.

Implementation of more playful experiences, in the form of educational mini-games, to raise awareness of fake news detection have also shown to improve participants’ ability to spot misinformation [3]. Even more, the development of playful educational tools that are available to the public have shown to attract participants in large scales, specifically by the teacher’s communities [19]. Which puts an effort in the design of tools that provide opportunities to educators to create lesson designs that adjust to their curriculum.

A common theme in the aforementioned studies is the creation of fictional use cases outside social media to raise awareness of potential threats and dangers. However, studies on promoting learning through social media need to be encouraged [11] and innovative approaches to provide interactive scenarios with hands-on experiences need to be formulated. Moreover, learning scenarios need to be designed without using censorship or restriction [48].

In this line, the introduction of initiatives to integrate educational opportunities within a controlled social media platform have been targeted as an opportunity. This integration would give the ability to educators to create educational materials attached to social media scenarios that would otherwise be difficult to be replicated in the classroom. To successfully achieve this, an educational component providing the resources to the students needs to be carefully designed to provide an engaging learning environment to the students and educate them towards a safer use of social media. This could be achieved with the implementation of a virtual learning companion (VLC).

A VLC can be characterized as a companion that shares a learning experience with the student and takes the role of a knowledgeable peer rather than appearing as an expert in the subject domain [45]. It usually takes the form of a computer-simulated character who engages in conversations with the student through an interactive chat interface [10].

The use of VLCs in educational scenarios is not a novel approach and it has previously shown its benefits towards students’ engagement and motivation [26]. Previous work has seen their successful implementation in the classroom in topics related to science [45] and mathematics [26]. Aside from cognitive skills acquisition, VLCs have also shown to be beneficial towards affective learning [56], and raising learning curiosity [57]. It is therefore evident that the use of VLCs in education can be beneficial across different domains. Work connecting the topic of SML and VLCs has seen the implementation of browser plugin digital companions, that provide individuals with tools to verify the credibility of digital content in social media [9] and support them in addressing misinformation through nudging mechanisms [33]. Moreover, VLCs can be traced to boosting [23], a particularly promising paradigm to increase online users’ competencies and counteract the potential drawbacks of social media. Boosts focus on interventions as an approach to improve people’s competence in making their own choices.

Further studies connecting diverse topics of SML and VLCs are scarcely available. With the aim to fill this gap in research, the COURAGE project proposes the use of VLCs towards the acquisition of SML skills. In particular, we are interested in investigating how VLCs can be used as part of SML interventions for the acquisition of diverse skills. Hence, seven empirical studies were designed each covering a different SML topic. In the following section, we present two different technical approaches of VLCs and then proceed to summarize seven diverse interventions that were designed. We then proceed to discuss the benefits and limitations of using this type of SML interventions.

4.1 COURAGE Browser Plugin

The first VLC was developed as a Chrome browser plugin to support users during social media content browsing. Once the user activates it, a chat-like interface appears on the right side of the webpage. Within the interface, a chatbot greets the users and guides them through learning material related to the content they interact with (see Fig. 1). During the interaction, the chatbot provokes users to reflect on content by asking questions, making suggestions and directing them to complete tasks to enhance their learning experience [2]. Initial work saw its implementation within a simulated social media environment that replicated Instagram (Pixelfed [39, 52]). This environment allows participants to interact between them by posting, commenting and liking content. Due to the nature of this work the environment was slightly modified to allow the implementation of educational material. Nevertheless its integration is also possible in real social media platforms.

The VLC interface. In this scenario, the student is able to interact with the companion and receive information related to images they have been seeing in their timeline.

Full size image

To evaluate the browser plugin VLC, the following empirical studies have been conducted or are currently ongoing within our own controlled social media environment, addressing the topics of fake news and everyday racism.

Fake News. Studies have shown that users with higher media literacy are more likely to identify fake news [27, 43], especially those with higher information literacy [28]. Therefore, interventions addressing the negative effects of fake news should focus on improving users’ information literacy skills, among others. In a combined study to evaluate the practicality of the browser plugin VLC and its effect to support learners in judging the veracity of content, a preliminary study was designed with participants aged from 19 to 65 years [1]. During the intervention, the participants were presented with three images that contained a fake, fact, and controversial item regarding COVID-19. Participants were asked to express their opinion about each image and classify it as fake or fact. It was also possible to give a neutral judgment (“not sure”). The companion then presented the users with links to the same images in different contexts based on reverse image search. This was supposed to help the participants in identifying fake news and encouraged them to explore external tools and resources to identify manipulated images. On their return, the participants were asked to reclassify the images and justify their choices. Results show that participants’ performance was improved after the interaction with the VLC. In a more recent study with the same environment and a larger user group, we could confirm that the agreement with experts ratings increased after the interaction with the VLC and that the judgments were more determined (less “not sure”).

Everyday Racism. Since racism also operates in everyday situations, for instance through incidental expressions [15], interventions targeting this form of discrimination are necessary. Similar to the scenario described above, young users between the ages of 12 and 19 explored a closed social media environment, receiving information from the VLC on the topic of racism related to an ad containing racism. For realizing this use case, however, InstaCour was used, which was also developed by the COURAGE group and represents a closed social media environment. In contrast to Pixelfed, students here do not have the option of adding comments under the posted images or liking posts. We intentionally chose this approach to keep the environment, and thus the conditions, consistent and uniform for all students throughout this iteration. We hypothesize that students who receive information on the impact and background of such everyday racism examples by the VLC will be more critical regarding social media content and more aware and literate regarding racism. Also, we suppose that users who interacted with the VLC and got information on racism will score higher in a quiz after the interaction than before. For the purpose of this study a social media ad was added to the closed network promoting a pen in “skin color,” prioritizing the interests of white people and thus reflecting white privilege and the dominance associated with it [40]. A full study design was developed, covering one experimental and one control group for testing the effects of the virtual learning companion, either transmitting information on racism (experimental group) or not (control group), on users’ knowledge and awareness regarding racism (behavioral and informational) in social media environments.

4.2 The Narrative Scripts

The next iteration of the VLC saw the implementation of the Narrative Scripts which combine elements of storytelling, collaborative learning activities and mini games [22] within an educational social media platform (ESM). The introduction of those elements within this context provided a further opportunity to enhance the educational material that students would receive as part of the VLC interaction. The incorporation of storytelling provides educators the ability to design lessons that can unfold within a controlled SM environment and create social media scenarios that would otherwise be difficult and dangerous to be replicated in the real world. The coordination of collaborative activities within the environment provides an opportunity to students to share opinions about the topics between them through role playing. Finally, mini-games serve as small activities or challenges that provoke students to reinforce the learning they received. The incorporation of these three elements within the VLC under the narrative scripts are complemented by the use of counter-narratives that can trigger different learning scenarios based on the students behavior in the ESM [35]. The use of counter-narratives in the topic of social media can be used to expose students to scenarios that counter the content they consume everyday. To investigate the narrative scripts as an educational tool towards the acquisition of SML skills two interventions were designed to explore the topics of social media self-protection skills and cyberbullying.

The narrative scripts interface. The student follows a story that is enriched by interactive learning components such as videos and multiple choice answers.

Full size image

Social Media Self-protection Skills (SMSPS). Whilst social media has its own sets of risks, adolescents might not avoid them as easily and may be less able to cope with difficult experiences. This could be due to key areas of the brain controlling decision making, self-awareness, and understanding of others not being fully developed. Consequently, putting them in greater risk of developing mental disorders [29]. Thus, we proposed the development of an intervention with the aim to teach adolescents how to cope with difficult experiences they might come across in social media. For this, an educational intervention was designed to evaluate the use of the narrative scripts to enhance adolescents’ SMSPS in the age range of 14 to 19. This intervention saw the design of six 2-hour sessions covering learning material related to the topic of digital self. During each session students followed the stories of four fictional teenagers and their experiences in social media through the VLC complemented by mini-games, collaborative activities, media (videos, images, voice etc.) and guided roaming. The stories covered different social media scenarios in topics such as wellbeing, reputations, social media and our brain, body image and AI algorithms.

Empathy Training. A second educational intervention proposes the design of a narrative script to sensitize adolescents to the negative effects of cyberbullying. We hypothesize that students who participate in the empathy training will show higher levels of perspective taking, higher intentions to help victims of cyberbullying, and lower intentions to reinforce bullies online, for example, compared to youth who do not participate in the empathy training. For this training, two videos were created in German and Spanish, containing examples and definitions of empathy and social media, respectively (see Fig. 2). In addition, two memes (i.e., an image enhanced with text) depicting bullying scenarios were created on which one of the authors is only partially visible. During the study, students will watch either the empathy video (experimental group) or the video about social media use (control group). In both conditions, the image is embedded as a posting in our own closed and controlled social media environment. Thereby, the students are guided by a virtual companion, either by being invited to empathize with the victims (experimental group) or by answering simply decoy questions regarding the social media video (control group). We plan to conduct the study with 12–15 year olds in Germany and Spain. In total, the completion of the pre and post questionnaires and the interaction with the virtual companion will take about 45 min.

4.3 Future Features Studies

In future studies, the VLC will be enriched with additional educational material and computing mechanisms aiming at further promoting and enhancing students’ SML skills. An initial step has been done in terms of evaluating the material and activities in schools (supported by teachers). Additionally, computing mechanisms have been explored as tools to further support students’ SML skills.

Emotional Intelligence and Fake News. Emotional Intelligence (EI) is described as the ability to identify and manage emotions, reason, decision making and problem solving. A recent study has shown that people with a high level of “EI” have fewer probabilities to be attracted by fake news thus showing a more conscious use of social media [46]. Therefore, to promote a more conscious use of social media, this study designed an educational intervention towards the acquisition and reinforcement of EI skills. In particular, it aimed to investigate the ability of adolescents to discern real or fake news in social media in relation to the levels of EI and verify the effect of an educational intervention to increase EI in adolescents. This study was aimed at adolescents aged from 16 to 19. Students were trained for ten 2-hour sessions in the following topics: awareness of emotions, empathy, assertiveness, interpersonal relationships, anger and impulse control, adaptability, problem solving, stress management, conflict management, and optimism. This study was conducted outside the VLC environment however a future implementation could foresee its adaptation in the VLC.

Image Manipulation and Unrealistic Beauty Standards. Social media can contribute to the normalization and trivialization of unrealistic and unhealthy beauty standards [24]. Being confronted with idealized and edited images of others can lead to a negative body image, driven by upward comparison processes [17] and may foster extreme impression management activities, again. Thus, it is critical for social media platforms to be transparent on the level of image editing that an image goes through. To enrich the VLC towards this matter, an image labeling feature was investigated towards its feasibility to be implemented as part of the ESM platform. The image labeling tool saw the tagging of images based on their level of alteration. An initial study examined how adolescents understood and accepted this approach [49]. For this, a mini game was developed exposing students to two different editions of a photo with descriptions underneath of the level of alteration in each. Students were then asked to select which image they prefer. After the mini game the students were asked to evaluate the feature. Findings have shown that image labeling can help to decrease affinity for edited images. Therefore, a future iteration would see the implementation of this tool within the VLCs.

4.3.1 Social Media Influence and Echo Chambers.

Finally, there is an importance to provide students with appropriate training to understand and deal with social media computing mechanisms and their possible drawbacks. Specifically students’ awareness of algorithm-based personalization of information, filter bubbles and the resulting social counterpart, echo-chambers, is crucial. An understanding of these computing mechanics would give students the basic skills and knowledge for recognizing and evaluating the impact of new technology in different situations and contexts [53]. Thus, an educational activity inspired by the “wisdom of the crowd” [38] theory was developed to raise awareness about the effects of social media influence and information personalization [5, 38]. The main objective of this work was to allow students to reflect on the effect of personalizing information to increase the awareness about the influence of social media. During the activity students received a talk related to the impact and differences of traditional and social media on user attitudes, beliefs and opinions [37]. The talk also covered topics related to personalisation algorithms highlighting the concepts of echo chambers and filter bubbles. Finally students were exposed to a “wisdom of the crowd” experience consisting in a social perceptual decision task, i.e. each user had to estimate how many dots were present in an image. In this experience the presence and bias direction of the social influence was manipulated by presenting aggregate estimations of all the participant choices that could be exact or biased, as only a polarized subset of users was selected. Presentation of biased estimates resulted in higher estimation errors. The students were later presented their performance and explained the presence of the two different conditions, exact and biased, as well as the parallel between the biased condition and the social media echo-chambers.

For most children and adolescents in the western world, social media adds a level of socialization. Youngsters interact within the social media sphere not only with the purpose to meet informational demands but also to construct citizenship. Compared to former generations, youth is fully oriented to digitalization as the material supports shaping their values and identity. Following Livingstone (2004), if it is true that teens spend an unbalanced time of active and passive communication, teens’ social media investment cannot be reduced to leisure and entertainment nor to the stigmatized image of “the media-controlling children”. Teens’ social media time is embedded in other socialization practices such as education, collaboration, and information-seeking [50]. Although banning access or restricting the time to social media seems to be a more efficient solution, prohibition will not cover the cost of lacking literacy. Providing social media situated scenarios such as learning about social media with social media might be an opportunity to grant access and security to young users. Just as people learn how to drive a car with a theoretical course and teacher-guided practice, training with tools such as VLCs can grant youngsters to safely navigate in virtual environments.

The VLCs developed as part of the COURAGE project may offer teachers, learners and families an enriching educational approach to social media. As part of the benefits observed in the different interventions, three opportunities have been listed to support the implementation of COURAGE VLCs for SML:

• Multiliteracy approach: VLCs can cover different areas of knowledge and topics. Although the tools have been designed to equip users with knowledge and competencies regarding the (mis)use and effects of information and technology, acquiring such competencies has an impact in other domain areas such as mental health and well-being. In line with scholars such as [7] and [50], COURAGE VLCs attempt to stimulate critical thinking from contextualized experiences. SML is produced as an outcome of the critical interpretation of scenarios where users access, use and apply information for daily living.

• Autonomy and self-guiding: VCLs have the potential to be deployed in school training without support from the developers in the form of educational tools. This facilitates its incorporation into formal school curricula or by external trainers (i.e., parents/tutors). Moreover, VLCs design makes users’ experience intuitive rather than instructional. The interactive experience, for instance with chatbots in narrative scripts, makes learning a gamifying experience. Consistent with self-determination theory, VLCs as a participatory support system also protect youth autonomy since the VLCs create choices instead of censoring content.

• Be Adaptive for learners: Similar to traditional social media platforms, VLCs can be automatized and supported by AI Algorithms. The benefits of implementing systems such as xAPI can guarantee tailored experiences for users to reinforce their learning dynamics.

However, the development and implementation of VLCs have their own set of critical points. To begin with, the replication of a social media platform and/or social media features may conduct users’ into more individual actions such as posting, liking, or sharing rather than interactive or collaborative practices. This could potentially become a threat to the learning aims of the material. It is therefore crucial that VLCs provide monitoring tools to administer the content and supervise online behavior. This leads to a more important discussion concerning the ethical issues and implications that the combination of social media platforms and educational tools might have (e.g., addressing young people whose identities are still developing, using potentially biased algorithms, deciding which behavior is acceptable and which not). This topic is addressed in more detail in [51].

Another important aspect to mention is the fact that social media is an environment constantly changing. The implementation of interventions for today’s popular media such as Instagram and TikTok may soon be outdated. The design of educational tools such as the VLCs may need to be updated accordingly to newer applications and youngsters’ interests.

Finally, despite the growing number of families and educators adept at technology and SML, others keep a restrictive approach toward social media. Some of the tools introduced in this paper were partially tested in schools (experiments finished earlier) as parents or teachers did not trust this SML approach considering the introduction of digital media and VLCs more inductive than reductive for social media use. Not until digital literacy would be considered part of children’s education, implementing VLCs (at least in schools) might be more complicated. Research evidence is, therefore, necessary to debate both the positive and negative outcomes of teaching with technology.

There is an immediate need for SML initiatives to educate about social media dangers through innovative educational methods. Hence, this paper introduced seven empirical studies that have been developed as part of SML interventions with the support of a VLC. Each study proposed a diverse solution covering different aspects of social media threats and dangers. Studies that have been conducted so far as part of this project have shown a positive impact of the VLC. In the case of the narrative scripts adolescents had a positive interaction with the educational activities generating intrinsic motivation [36]. Moreover, students using the browser plugin to reflect on fake news veracity have shown more deterministic judgments after interacting with the VLC [1]. Finally, students engaging with a scenario powered by machine learning mechanisms have shown an increase in their understanding of how such algorithms work [37]. With a few more studies currently in the line to form our final judgment, we believe this work has paved the way for future research combining these two fields.

However, we believe that SML interventions cannot be taken as the only solution to all the problems surrounding the world of social media. It is also necessary to encourage teachers, parents and caregivers to participate in training to support schools with regard to knowledge and awareness about SML. Overarching interventions are needed to cover users’ social cultural backgrounds.