Literature review

A number of studies have looked at the role of gender to understand disparities in internet use and digital skills. Although many of these studies have been conducted on limited samples of high school or college students, reviews that also include adult samples indicate that men appear to be more likely to exhibit problematic internet use than women (Baloglu et al., 2020; Morahan-Martin, 1998). Men and women also engage in digital environments for different reasons. When focusing on youth, we also find gender differences in the ways they use the Internet (Herring & Kapidzic, 2015). Livingstone and Bovill (1999) showed in the late 20^th century that boys used computers more often than girls and felt more comfortable doing so. However, these gender differences disappeared quickly, and by 2004, boys and girls utilized the internet equally to communicate with peers. In the late 2000’s, girls surpassed boys as the most frequent internet users, largely fuelled by the growth of social media platforms (Lenhart et al., 2007). These new media were – and continue to be – more commonly used by girls than boys to communicate with peers and to create and share videos (Lenhart, 2012). Boys were more likely to use these technologies for entertainment (e.g., gaming) and visit video websites like YouTube (Rideout et al., 2010).

The digital development of non-binary youth has not received much attention in Europe. In the United States, evidence from 2013 shows that LGBTQ youth spend an average of 5 hours per day online; approximately 45 minutes more than reported by non-LGBTQ youth (GLSEN et al., 2013; Hatchel et al., 2021b). This is not entirely surprising: The Internet (and, more specifically, social media) has quickly evolved into an arena that provides LGBTQ youth an opportunity to safely construct and develop their sexual and gender identity, interact with people from their community, and establish connections with likeminded individuals – all of which may be absent in their face-to-face relationships (Lucero, 2017). A recent large-scale survey study in the United States and Canada confirms that LGBTQ youth are highly active on the Internet and report high usage of new ICTs (McInroy et al., 2019a). Additionally, it also reports that they are more likely to participate in online than offline LGBTQ communities due to heightened feelings of safety and increased emotional support (McInroy et al., 2019b). However, this greater internet use is also related to a variety of (digital) risks, most notably through cyberbullying, misinformation, sharing and receiving sexually explicit images, and meeting up with online contacts (Hatchel et al., 2021b; Sousa et al., 2020). Recent data indicates that about one-third of LGBTQ youth report being a victim of cyberbullying, either due to their gender identity or their sexual orientation. Varjas et al. (2013) found that those who reported cyberbullying victimization also had a greater likelihood of being victimized by face-to-face bullying. Regarding misinformation, it has been reported that online resources and social media that were regularly consulted by transgender youth (e.g., Tumblr) for health concerns or LGBTQ rights often contain considerable misinformation regarding these topics. While this is problematic for all youth, this is particularly the case for LGBTQ youth given their reliance on online resources for psychoeducation about their gender or sexual development (Hatchel et al., 2021b).

Co-rumination is a final risk that we review here. It is defined as “excessively discussing personal problems within a dyadic relationship and is characterized by frequently discussing problems, discussing the same problem repeatedly, mutual encouragement of discussing problems, speculating about problems, and focusing on negative feelings” (Rose, 2002: 1830), and it is considered common among LGBTQ youth in digital settings. Although outcomes of co-rumination are not exclusively negative – it has been known to increase friendship quality –, the normalization of mental health problems like depression, anxiety, and suicide that result from it outweigh its positive outcomes (Meyer et al., 2015). These risks – along with other factors – are linked to poor mental health, psychological distress, and suicidal ideation among LGBTQ youth (Hatchel et al., 2021b), with rates that are considerably higher than those of non-LGBTQ youth. Given these youth’s reliance on digital environments and the risks that are associated with their frequent use of these sources, more research into protective factors against such risks is pivotal among LGBTQ youth. In this regard, various studies have highlighted the role of digital skills (Livingstone et al., 2021; Rodríguez-de-Dios & Igartua, 2016). Youth is expected to be able to avoid the negative outcomes of digital technologies by acquiring digital skills (Rodríguez-de-Dios & Igartua, 2016). However, recent evidence suggests that the relationship between risks and skills is not straightforward. In their systematic evidence review, Livingstone et al. (2021) showed that digital skills were indirectly linked to greater exposure to digital risks. They also found a positive link between skills and online opportunities, information benefits (of particular importance to LGBTQ youth), and orientation to technology.

Regarding Livingstone et al.’s (2021) conflicting results on the relationship between digital skills and digital risks, they did find that specific subsets of skills were linked to lower exposure to digital risks. For example, technical skills were linked to mixed or even negative outcomes, while information skills were linked to positive outcomes (Livingstone et al., 2021). Contemporary understandings of digital skills dictate that it is a multidimensional concept, consisting of five subdimensions: (1) technical/operational skills; (2) programming; (3); information navigation; (4) communication and interaction; (5) content creation and production (Helsper et al., 2020). These various skills are linked to outcomes in different ways, as illustrated by Livingstone et al. (2021). In addition to outcomes, a number of key antecedents of digital skills have also been identified. An important hypothesis here is the recursive loop hypothesis which suggests that specific socio-psychological and structural factors may reinforce digital inequalities and disparities, which, in turn, negatively affect these factors (Robinson et al., 2020). We focus on the role of peer and family support and self-efficacy. Peer and familial support structures have been found to be of key importance to LGBTQ youth’s well-being (Fish et al., 2020; Hatchel et al., 2017), but also to the development of digital skills among youth in general (Mascheroni et al., 2022).

The present study

This article provides a first investigation of the development of digital skills and digital activities of non-binary youth, comparing these with those of binary youth. Currently, there is a dearth of research on non-binary youth (McInroy et al., 2019a). The few studies that do exist focus mostly on mental or physical health outcomes of these youth. However, as the LGBTQ community grows, it becomes increasingly important to include specific studies on these youths in other areas of social and psychological sciences that are not specifically related to their gender or sexual identity. Not doing so would yield an incomplete picture of the social reality of contemporary societies, particularly for young people.

In this study, we focus on the digital development of non-binary youths. This is directly relevant to this group, as studies have shown they spend significantly more time online and use the internet differently than boys and girls, while also being confronted with a number of digital risks which negatively contribute to their mental health (McInroy et al., 2019a, 2019b). Digital skills are hypothesized to be a key protective factor against such digital risks. Our expectations are twofold. On the one hand, given that LGBTQ youth spend more time online than non-LGBTQ youth (McInroy et al., 2019a), we may expect that their digital skills are more developed. On the other hand, a lack of digital skills may be one of the reasons why these youths are frequently confronted with digital risks. In sum, we seek to answer the following research question:

• RQ1. How do digital skills and digital activities differ between non-binary and binary youth?

With regard to mental health and internet use, more empirical evidence exists – particularly from studies in the United States. LGBTQ youth are known to report high rates of poor mental health, psychological distress, and suicidal ideation (Hatchel et al., 2021b). It is unclear to what extent these adverse mental health outcomes are linked to their internet use, but as well, evidence indicates that LGBTQ youth spend more time online than non-LGBTQ youth – up to 45 minutes per day more on average (GLSEN et al., 2013). To confirm these more well-known insights for the current sample, we test two hypotheses:

1) Non-binary youth report greater internet use than binary youths (Hypothesis 1).

2) Non-binary youth report lower well-being than binary youth (Hypothesis 2).

Because the literature on non-binary youth is scant, we are unable to develop clear hypotheses for the link between well-being and internet use for this group. Thus, we also develop an additional research question regarding the link between well-being and internet use:

• RQ2: How is well-being linked to internet use among non-binary youth?

In line with the recursive loop hypothesis, we believe that digital skills and digital activities may be reinforced by certain socio-psychological and structural factors. Three factors that we examine are peer support, family support, and self-efficacy because of their relevance in both, digital skills and LGBTQ literature (Mascheroni et al., 2022). Because it is unclear to what extent these factors are related to digital skills and digital activities among non-binary youth, a third research question is:

• RQ3. How are digital skills and digital activities linked to support structures and self-efficacy?

Data

We distributed an online questionnaire to children and young people aged 11 to 20 in six European countries (Estonia, Finland, Germany, Italy, Poland, and Portugal) between April and November 2021 (N=6,221) ² . The data were collected in collaboration with secondary schools. We used a convenience sampling design, although schools were selected based on their socioeconomic status to ensure the diversity of participants. We contacted and informed schools about the study. A total of 52 schools agreed to participate, with 460 classes participating in total. Researchers were present during the data collection in each class, either in person or digitally, due to school closings during the COVID-19 pandemic. Informed consent was obtained from all participants and their legal guardians. The questionnaire was identical across all countries and it was presented in the official language of each country. Translations were carried out by professionals.

Measures

• Gender: To assess respondents’ gender identity, we asked to what extent they identified as: (1) boy, (2) girl, (3) non-binary.

• Digital skills: We used the youth Digital Skills Indicator (yDSI), a validated cross-cultural scale that measures five dimensions of digital skills: (1) technical and operational skills; (2) programming; (3) information navigation and processing; (4) communication and interaction; (5) content creation and production. It was developed to capture key elements of functional (ability to use ICTs) and critical (understanding ways how ICTs are designed, and content is produced) digital skills (Helsper et al., 2020); 25 items were used to measure digital skills. A sample item was ‘I know how to turn off the location settings on mobile devices’, and answer options ranged from 1 (not at all true of me) to 5 (very true of me). Programming was measured through a single item, the other skill dimensions, each measured using six items, had high reliability in the current sample (technical and operational: .74; information navigation and processing: .80; communication and interaction: .77; content creation and production: .79). The full scale can be found in Helsper et al. (2020).

• Digital activities: We presented participants with 11 items to assess five different types of digital activities: (1) online learning; (2) social relationships; (3) entertainment; (4) content creation; (5) health use. A sample item was ‘I used the internet to search or follow news about local, social, environmental, or political issues’, and answer options ranged from 1 (never) to 6 (almost all the time). The specific wording of the items can be found in Table 1.

• Internet use: We included an indicator regarding the time spent on the Internet during a regular weekday, with answer options ranging from 1 (little to no time) to 9 (about 7 hours or more).

• Well-being: Well-being was assessed through six items that asked whether participants felt happy, pleased with the way they were, felt that life was enjoyable (positive dimension), felt dissatisfied with life, felt cheerless, and felt that life was meaningless (negative dimension), in the past year. Answer options ranged from 1 (never) to 4 (often). These six items were aggregated into a positive dimension and a negative dimension.

• Peer support and family support: Peer and family support were each measured through three items. Items that assessed peer support were ‘My friends really try to help me’, ‘I can count on my friends when things go wrong’, ‘I can talk about my problems with my friends’. Items that assessed family support were ‘When I speak someone listens to what I say’, ‘My family really tries to help me’, and ‘I feel safe at home’. For these items, answer options ranged from 1 (not true) to 4 (very true).

• Self-efficacy: This construct was measured through a four-item scale developed by Schwarzer and Jerusalem (1995). A sample item was: ‘It’s easy for me to stick to my aims and achieve my goals’. The response scale ranged from 1 (not true) to 4 (very true).An overview of the full sample can be found below, and the sample by country can be found in Table 2.

Data analysis

We conducted a Pearson correlation analysis between relevant variables for non-binary youth, and boys and girls (Table 3). In this and the following analyses, we chose not to distinguish by country for two reasons. First, as argued by Gui and Argentin (2011: 964), “among today’s teenagers in Western countries differences in terms of physical access [to the Internet] are almost irrelevant. Nowadays, schools are increasingly offering an internet connection so that access is free and easily available for many high school and college students. In some areas the binary divide between the haves and have-nots no longer applies to young people”. To further support this, Eurostat (2020) indicates that 94% of young people in the EU-27 made daily use of the internet in 2019. Second, the share of non-binary youth was low (under 4%) in all countries. Splitting the analysis by country would further fragment the limited share of non-binary youth which would endanger the robustness of findings of country differences in these analyses. Furthermore, no non-binary youth were included in the Portuguese sample.

In the first step, we ran a one-way analysis of variance to study whether digital skills, activities, internet use, and well-being significantly differed between boys (n=3,119), girls (n=2,991), and non-binary youth (n=106). This analysis provides an answer to the main research question of how digital skills and activities differ between binary and non-binary youth, and to the hypotheses regarding well-being (H1) and internet use (H2). Subsequently, we ran ten linear regression analyses with the five digital skills and five digital activities as dependent variables, and selected socio-psychological and support variables as independent variables to provide an answer to RQ2. To reduce the odds of Type II error resulting from the small sample size for non-binary youth (n=106), we applied a 1,000-sample bootstrapping method. This has been shown to be an effective statistical procedure that reduces the standard error of parameters under the condition requiring both acceptable confidence intervals and confidence levels (Chernick, 2011).