The rise of social media has democratized information sharing but also facilitated the spread of harmful and illegal content. Social media platforms have established community guidelines and content moderation policies to address this, leading to a debate about the legitimacy of private regulation of fundamental rights like freedom of expression. While some focus on legal and human rights frameworks, this study takes an alternative approach, examining public perceptions of the moral credibility of these norms and the extent to which the public supports content removal based on community guidelines. The research questions center on whether a consensus exists on rules limiting VHC on social media and what factors influence this consensus. Prior research suggests generally favorable attitudes towards regulations, but the perceived legitimacy and influencing factors remain unclear. This study uses an experimental design to investigate these questions focusing specifically on VHC.

Existing literature highlights the challenges of regulating online hate speech, as legal frameworks vary across countries and legal traditions. The suspension of Donald Trump's social media accounts following the January 6th Capitol attack exemplifies the complexities of this issue, raising questions about the legitimacy of private companies limiting free expression. While some argue for legitimacy through rule-of-law principles, this study adopts a different approach by examining public perceptions of the morality and legitimacy of community guidelines. The study draws upon criminological literature emphasizing the importance of perceived legitimacy for voluntary compliance with norms and sanctions. Previous studies suggest public support for platform regulation, but findings regarding the perceived fairness and transparency of the processes remain mixed, with concerns raised about the limitation of freedom of expression. Prior experimental studies on crime seriousness reveal the importance of using concrete examples and highlight factors such as harmfulness, wrongfulness, and frequency in shaping perceptions.

The study employed an experimental between-group design with three independent samples of approximately 300 Spanish participants each, recruited through non-probability snowball sampling. Participants were presented with 13 descriptions of discursive behaviors extracted from Twitter's rules, representing a common substrate for major platforms (Twitter, Facebook, Instagram). Three experimental conditions varied the presentation style: 1) description only, 2) description and example, 3) example only. Participants rated each behavior based on six factors: a) moral reproach (wrongfulness), b) risk/harm (harmfulness), c) frequency, d) overall seriousness, and e) agreement with content removal. A 0-10 scale was used for most ratings, except for frequency (0-5 Likert scale). Socio-demographic data (age, gender, education, legal education, political ideology) were also collected. Data analysis involved comparisons of means and standard deviations across groups, ANOVA, t-tests, and linear regression modeling to assess the relationships between variables.

The study's key findings demonstrate a high degree of consensus on the seriousness of VHC and the need for content removal. However, the introduction of concrete examples increased the variability in ratings, reducing consensus, particularly in situations where only an example was provided. This suggests that while abstract concepts related to VHC elicit broader agreement, specific examples may lead to more nuanced and varied judgments. The results also indicate that behaviors involving direct physical harm are perceived as more serious than those involving only moral harm. This difference was amplified further when examining VHC directed at vulnerable groups (hate speech) versus those targeting the general population, highlighting societal concern around hate speech. Contrary to previous research, the study found that perceived harmfulness is a stronger predictor of seriousness than perceived wrongfulness, emphasizing the practical implications of online discourse in the physical world. Linear regression analysis revealed a significant positive correlation between perceived seriousness and support for content removal (approximately 31% of the variance explained), highlighting the link between these two assessments. The study further revealed that harmfulness is a key driver of seriousness, highlighting a concern over potential real-world consequences stemming from social media discourse.

The findings confirm high social concern about VHC on social media and support for content removal policies. The reduction in consensus when specific examples are presented suggests a potential need for greater transparency and clarity in community guidelines. The emphasis on harmfulness as a primary predictor of seriousness indicates a concern regarding the real-world impact of online hate speech, and supports the claims of companies like Meta and Twitter that justify their policies based on risks to the physical world. While the correlation between perceived seriousness and support for content removal reinforces the general public's acceptance of platform regulations, the model only explains 31% of the variance. This calls for further exploration of other factors that may influence individuals’ responses to online VHC and enforcement of platform rules, focusing on the procedures for moderation and their impacts on individuals' perception of justice and free speech..

The study shows significant consensus on the seriousness of VHC and the need for content moderation on social media, even though concrete examples can reduce consensus. Harmfulness emerged as a stronger predictor of seriousness than wrongfulness, highlighting public concern for real-world consequences. While there is a strong correlation between perceived seriousness and support for content removal, it's important to study factors influencing support and address concerns about freedom of expression. Future research could explore the impact of different content moderation mechanisms, user experience, and transparency on the perceived legitimacy of platform rules.

The study's use of a non-probability snowball sampling method limits the generalizability of findings to the wider population. The sample's overrepresentation of women might have influenced the perceived seriousness scores. The focus on Twitter's community guidelines might not fully capture the nuances of other platforms' policies. Finally, the study only explores a specific aspect of platform regulation – public support; it omits crucial considerations of the procedural mechanisms and enforcement processes in determining the broader legitimacy of such regulations.