Background
With the rise of digital platforms, the management of hate speech and online harms has become an important topic in their governance. Many people believe that the biggest challenge in managing hate speech is how to define it.
However, research has found that the management of hate speech is also related to various factors such as platform power, algorithm distribution, technical identification capabilities, reporting mechanisms, and accountability. More importantly, the governance approach of deleting posts is not sufficient to fully address this structural issue.
This phenomenon allows for the analysis of a dimensional subset of the aforementioned issues. Lao A, after whooping attention for provocational narratives such as “American Killing Line”, Lao A egoistically and sexually humiliates in his live streaming and online discourses, addressing in a monolithic fashion the Chinese female international students and their mother chaperones. He further extends unverified personal gossip into group moral accusations (Wangyi, 2026).
The primary comments are reported to be the origin of an avalanche of comments in subsequent posts, re-posting, and multiple acts of dissemination, which eventually led to social stigmatization and an online enclosure of the female international students.
This example helps to better highlight and precisely articulate what the problem of governance of hate speech is. It is not simply a matter of identifying the content of the speech, but it is much more concerned with the ways in which online platforms provide certain features to boost visibility, the ways users choose to stigmatize, and finally, how users deflect and shift the blame after the damage is done.
- The dilemma of definition and power structure
The challenges and difficulties faced in the governance of hate speech are diverse in practical applications. Firstly, it is difficult to establish a clear and definite boundary for defining hate speech. There is still no universally accepted definition, which means that the expression and perception mechanisms of hatred are still a topic that needs further research (Assimakopoulos et al., 2017, p.3). There is literature indicating that there are significant differences between the current legal understanding of this term and the multiple and covert forms of hate expression in reality (Assimakopoulos et al., 2017, p.4).
As a result, there is a considerable amount of language with derogatory, defamatory, and abusive connotations that may not be legally classified as hate speech, but they can still have serious negative effects on recipients through mental harassment (Assimakopoulos et al., 2017, p.4).
However, some opponents of hate speech laws may question whether these harmful acts are direct enough or serious enough to justify limiting freedom of expression (Gelber & McNamara, 2015, p.325). Therefore, the difficulty in governing hate speech is not only due to the lack of a unified definition. There is a considerable vacuum zone that is difficult to overlap between legally identifiable forms of harm, empirically perceptible forms of harm, and forms of harm that governance institutions are willing to recognize.
In addition, the issue of defining hate speech cannot be simply regarded as a dilemma in terms of definition, as there are also controversies arising from power structures in the governance process. Some have pointed out that although various digital platforms today increasingly resemble public infrastructure, their essence remains that of private companies (Flew 2021, p.99-100). This means that there is a clear gap in the existing regulatory framework for hate speech, as it heavily relies on platforms for self-management (Flew 2021, p.99-100).
Platforms can determine the definition of certain hate speech and still hold governance power as a private entity, even if it is recognized by the public as a public infrastructure. In this way, the problem is no longer just how to define the content, but who decides which expressions can continue to exist, which expressions should be removed, which groups will be prioritized for protection, and who will ultimately bear the consequences of judgment errors.
Lao A’s incident precisely illustrates this point. Regarding this incident, public opinion can continue to debate on the surface whether these remarks constitute strict hate speech, or whether they are just extreme expressions. Because his expression during the live broadcast seemed to be just a discussion. The fundamental problem with this statement lies in his promotion of unverified cases and sexually derogatory remarks to the overall judgment of the group of female international students and accompanying mothers. In this way, the focus of discussion shifts from a fact to moral skepticism and identity stigma towards the entire female population.
But before the definition dispute is resolved, the harm has already fallen on the targeted group first. Female international students and accompanying mothers’ social media accounts may be insulted, suspected, asked to prove their innocence in the comment section, and even automatically attributed to some negative imagination without any specific evidence. They have already borne the direct consequences of the spread of stigma.
Moreover, in this process, the responsibilities between the platform, internet celebrities, and bystanders were dispersed, and the initial framework of stigma was created by the speaker. The platform uses hot search, recommendation, and high interaction mechanisms to increase the visibility of these comments. Users in the comment section continue to copy and strengthen this framework, forming an encirclement of the group. In this chain, harm is not caused by a single individual, but by multiple parties working together, but ultimately the labeled female group bears the consequences.
- Algorithm, visibility, and structural damage
As more and more people stay, communicate, and obtain information online, the rules and data processing methods behind algorithms have begun to influence what people see, pay attention to, and how they will think and do next (Flew, 2021, p.108-109). Therefore, attention allocation on the internet is increasingly not naturally formed, but arranged by platforms that rely on algorithms to filter content. This indicates that algorithms are not just tools for platforms to process information, they also determine which content is easier to see and which topics are easier to magnify.
Additionally, studies have found that content containing hate speech spreads faster, has a wider scope, and is more likely to be seen by more people compared to ordinary content (Mathew et al., 2019, p.173). Moreover, the connections between these hate users are also closer (Mathew et al., 2019, p.173). This indicates that the impact of hate content is not only determined by the content itself, but also by how quickly it spreads on the internet, how widely it spreads, and whether a closer dissemination network is formed between publishers.
In other words, online hatred is not always an individual expression that appears sporadically and ends after being published but is constantly spread and strengthened in a more closely connected and rapidly spreading network. Therefore, the damage it causes is also more likely to persist and has more obvious structural characteristics.
This phenomenon is closely related to the development of Internet media characteristics. Research has found that these features of the World Wide Web provide the foundation for the growth and rapid spread of online hate phenomena (Assimakopoulos et al., 2017, p.11-12). The internet provides conditions that facilitate the spread of hate speech, including rapid circulation, easy participation, anonymity, and regulatory difficulty (Banks, 2010, p.233).
Lao A’s incident can help illustrate how this structural damage is formed. If these comments were only made on one live broadcast, their impact could have been relatively limited, at most just a controversial personal expression. However, in the process of platform dissemination, these statements did not remain in the original context, but were constantly captured, forwarded, edited, and retold, and compressed into more easily spreadable and derogatory labels in the comment section (Weizhou, 2026). As these expressions move away from the initial live streaming scene, they are no longer just a single description of a phenomenon by a speaker but gradually transform into a stigmatizing framework that can be repeatedly invoked, replicated, and reinforced by many users.
In this process, the cause of online harm also changed. The actual growth of harm is no longer determined by the original utterance of Lao A, but the product of the joint action of platform popularity mechanism, comment mechanism, and secondary dissemination logic. The controversial and provocative statements are more readily noticed by the heat mechanism, and the comment section offers the space of the centralized reproduction of the stigmatized language. The secondary modes of dissemination like screenshots, reposts and brief video clips also make these expressions even more aggressive and condensed.
In this way, the claims that originally revolved around certain unverified cases have been constantly generalized into an overall imagination of the group of female international students and accompanying mothers, ultimately leading to individual remarks evolving into online witch hunts targeting specific female groups.
- Limitations of deleting posts and banning accounts
Currently, hate speech is dealt with in a rudimentary way and is most addressed through post removal and banning accounts. However, these ways are not an effective regulatory mechanism, as stated before, the mechanism of harm embedded in hate speech is not just in discourse, but in the recommendation system, visibility, interactivity, incentives, appeals, and the very structure of responsibility of the platform. For this reason, governance in a linear way as “harmful content is discovered and then it is removed” is not sufficient to grasp the phenomenon of online hate in its complexities.
The reason why deleting posts or banning accounts is not enough is that the platforms are not just a passive place to store content. It is a technological environment that allocates content visibility and attention flow direction (Flew, 2021, p.108-109). The platform needs to attract users to continue using it for its own benefit, and the attraction brought by hate speech is very significant, as it can easily trigger different emotions and reactions from people.
In such circumstances, the development of online harm does not merely depend on the content itself. Since even in the case when the same content is inserted into various systems of recommendation and visibility distribution, the effect of its dissemination and the consequent harm effects will be utterly different. Therefore, when governance remains at the stage of the deletion of the posts, it will not be able to address the structural risks which the platform system introduces.
Lao A’s incident exemplifies the weakness of post deletion governance. Although the platform may block the original account, live content, or other posts following the spread of the controversy, stigmatized language may be disseminated through screenshots, paraphrasing, comments, and secondary broadcast. As the platform starts to intervene, the harm typically stops existing in the original content source, but moves into a more diffused and challenging to locate dissemination form. The original speech can be erased, but the tags, hints and moral imaginations based on it will not automatically vanish and they will still be spread in new posts, new accounts and other commentaries.
This indicates that deleting posts or banning accounts often only deals with the most intuitive and visible content node, rather than the entire mechanism of the actual occurrence and continuous spread of damage. Once insulting labels and moral skepticism enter broader online discussions, they gradually detach from the original context and become collective expression resources that can be repeatedly invoked. At this point, many users can still use the widely circulated stigma framework to mock, attack, and judge female international students and accompanying mothers, even if they no longer directly quote the original expression of Lao A. So, the continuation of harm no longer depends on whether specific content still exists, but on whether stigma has been embedded in a broader interactive order.
In this sense, post deletion governance has obvious retrospective and superficial characteristics. It can delete a post or restrict an account, but it is difficult to reverse the already formed group humiliation environment. It is also difficult to deal with the repeated hostility in the comment section, the amplification effect of algorithm recommendations on controversial content, and the continuous reinforcement of stigma by onlookers in interaction.
Conclusion
In conclusion, the difficulty of governing hate speech cannot be simply seen as a purely defining issue, as there are always power, responsibility, and political judgments behind the controversy over definitions. Insufficient platform algorithms, anonymity, cross-border dissemination, and automated auditing make online harm more persistent and structural.
The incident in Lao A further illustrates that stigmatized remarks targeting specific groups can be quickly spread through platforms and transformed into broader online campaigns, and this harm cannot be completely resolved by deleting a single piece of content. Therefore, the governance of hate speech and online harm should not only focus on the content itself, but also extend to the distribution logic, interactive order, and responsibility structure of the platform.
References
Assimakopoulos, S., Baider, F. H., & Millar, S. (2017). Online Hate Speech in the European Union. In SpringerBriefs in Linguistics. Springer International Publishing. https://doi.org/10.1007/978-3-319-72604-5
Banks, J. (2010). Regulating hate speech online. International Review of Law, Computers & Technology, 24(3), 233–239. https://doi.org/10.1080/13600869.2010.522323
Flew, T. (2021). Regulating Platforms. Polity Press.
Gelber, K., & McNamara, L. (2015). Evidencing the harms of hate speech. Social Identities, 22(3), 324–341. https://doi.org/10.1080/13504630.2015.1128810
Mathew, B., Dutt, R., Goyal, P., & Mukherjee, A. (2019). Spread of Hate Speech in Online Social Media. Proceedings of the 10th ACM Conference on Web Science – WebSci ’19, 173–182. https://doi.org/10.1145/3292522.3326034
Wangyi. (2026). 牢A直播聊女留学生和陪读妈妈 个别行为引发群体标签争议_新闻频道_中华网. China.com. https://news.china.com/socialgd/10000169/20260123/49189483.html
Weizhou. (2026, January 30). 冰川思享号｜炮制“斩杀线”的牢A，为何要造留美女生黄谣. 中国数字时代. https://chinadigitaltimes.net/chinese/724803.html
Be the first to comment