Background: Online hate speech (OHS) is a complex social problem, but many studies only focus on perpetrators. Therefore, we conducted a pre-registered online survey that also examined victim and bystander variables. The survey was based on an action-theoretical model of OHS (Author, 2023) which synthesized established aggression theories.

Research questions: Among other research questions, we examined which values and attitudes predict whether bystanders behave passively (RQ5), support the victim (RQ6a; RQ6b), or support the perpetrator (RQ7).

Method: We sampled a national quota sample of Internet users in Germany (N = 1,791; Mage = 44.1, 48.3% women) and calculated a logistic regression for each of the four RQs with age, gender and education as covariates.

Predictors were “experience with offline bullying as a bystander” (α = .88; Pfaff et al., 2007), “wanting to be a role model for others” (α = .73; Heinitz & Rowold, 2007), “sensitivity to injustice” (as a victim α = .73, as a bystander α = .61, as a beneficiary α = .66 and as a perpetrator α = .54; Beierlein et al., 2013), “compassion” (α = .85; Hwang et al., 2008), “social responsibility” (α = .85; Bierhoff, 2000), “moral courage” (α = .86; Kastenmüller et al., 2007) and Schwartz's values (“self-enhancement” α = .74, “openness to change” α = .69, “self-transcendence” α = .82 and “conservation” α = .66; Boer, 2013).

To assess the criterion variables, respondents were shown five different vignettes and asked whether and how often they had seen this type of OHS but otherwise done nothing (RQ5 passive bystander), supported someone who received this OHS (victim-supporting bystander with variants RQ6a “dislike” and RQ6b “report”), or supported someone who produced this OHS (RQ7 perpetrator-supporting bystander). The five vignettes were combined into ordinal scales (all α > .91), which were transformed into binary scales to improve the model fit.

Adolescents are increasingly exposed to cyberhate—online hate speech and biased content (Kardefelt Winther et al., 2023; Machackova et al., 2020), raising concerns about its long-term psychological and behavioural consequences. While existing research highlights the harmful effects (Jaron Bedrosova et al., 2024), the processes through which frequent exposure influences psychological and behavioural outcomes remain underexplored. To address this gap, we adopt desensitisation and social learning frameworks (Allen et al., 2018; Anderson & Bushman, 2002; Bandura, 1978) to shed more light on these mechanisms.

Prior research with adults links short-term cyberhate exposure to desensitisation (Soral et al., 2018, 2023), but long-term effects and effects on adolescents are less understood. Yet adolescents’ identity and intergroup attitudes are developing (Cortese, 2005) and may be shaped by frequent exposure to cyberhate’s biased messages. Testing the desensitisation assumption, we investigate whether frequent cyberhate exposure increases adolescents’ moral disengagement (i.e., minimising consequences of cyberaggressive incidents). Further, youth cyberhate exposure has been cross-sectionally linked to perpetration (Wachs & Wright, 2018; Wachs et al., 2021). Adopting the social learning theory, we investigate whether frequent exposure fosters learned cyberhate behaviours in the form of cyberhate aggression. We also explore whether moral disengagement mediates this association.

We will use longitudinal online survey data from 3,087 Czech adolescents (ages 11-16, M=13.47, SD=1.74; 50.1% boys) collected over four waves, six months apart, in 2021-2022. We used quota sampling to ensure (1) that included households represent Czech households with children in terms of SES, region, and municipality size, and (2) balanced age and gender groups. We measured cyberhate exposure and perpetration using two single-item measures and moral disengagement using a four-item scale (adapted from Garland et al., 2016). We will test within-person longitudinal effects between these variables using the random intercept cross-lagged panel model (Hamaker et al., 2015).

Although social media may foster civic engagement and facilitate the exchange of diverse opinions and perspectives, exposure to socially harmful content such as digital hate has become commonplace even within mainstream platforms. Across Europe, digital hate is an escalating concern, with women among the most frequently targeted groups. Frequent exposure to such content comes with negative consequences, including psychosocial impairment, diminished participation in online discourse, and sometimes even threats to personal safety. To develop effective, trans-European countermeasures, it is essential to identify dispositional and attitudinal factors that are associated with user engagement against sexist digital hate and support for platform-led moderation efforts, while accounting for varying socio-political contexts. To address this, we conducted a cross-sectional survey in Austria (n = 791), France (n = 800), Hungary (n = 775), and Sweden (n = 785)—countries varying in their regulatory approaches and law enforcement in combating digital hate—drawing from quota-representative (based on age, gender, and education) general population samples of social media users, and thus capturing broader societal patterns that may shape digital hate dynamics across Europe. Using multi-group path analysis, we examined how online self-presentation, toxic disinhibition (both selected via a recently proposed process framework of self-disclosure), empathic perspective-taking, empathic anger (chosen as distinct action-facilitating traits derived from appraisal theory), and misogynistic attitudes (selected due to topical relevance) relate to direct and indirect intervention activities, as well as support for content moderation (while exploring how gender may moderate relationships). Additionally, within the female subsamples, we investigated whether recent victimization experiences with sexist digital content influence these associations. Altogether, this study enhances our understanding of user responses to sexist digital hate with the goal of contributing to evidence-based strategies for mitigating hateful communication across Europe and, ultimately informing policy initiatives for safer digital environments that foster proactive user engagement.

Recently, considerable efforts have been made to investigate the factors that influence bystander behavior in the context of cyberbullying on social media. Most of the past research studied bystanders’ personal-level variables and contextual factors of the incident. However, there is a lack of attention on investigating the potential impact of cyber-victims’ help-seeking behavior on bystanders’ behavioral intentions. According to an earlier study on the effectiveness of cyber-victims’ coping strategies (Machackova et al., 2013), active help-seeking has been recommended as an effective emotional coping strategy and has the potential to mitigate cyberbullying. Previous studies have also shown that cyber-bystanders tend to prefer offering private support over public support. Building on these findings, we tested two pre-registered hypotheses in an online experiment: (1) Cyber-bystanders are more likely to intervene if they receive an explicit request for help than if they receive no request for help. (2) Cyber-bystanders are more likely to intervene if they are being asked for help privately than if they are asked for help in public. Participants (N = 152) aged 18 to 30 were randomly assigned to one of three between-subject conditions. In each condition, they took the perspective of a cyber-bystander and saw a screenshot of a fictitious cyberbullying scenario in a chat room in which the cyber-victim either did not ask for help (control group), asked for help in a private chat (experimental condition: private help-seeking) or publicly in the chat room (experimental condition: public help-seeking). As dependent variables, participants indicated their intentions to help the victim, to remain passive, and/or to attack the victim. In addition, they completed an empathy questionnaire. Participants’ data on gender, age, education level, social media use, recollection of help request, interpretation and perceived severity of the scenario were also collected for further explorations. Data was analyzed as preregistered.