（https://www.theguardian.com/technology/2022/oct/28/elon-musk-twitter-hate-speech-concerns-stock-exchange-deal）
In recent years, digital platforms have increasingly presented themselves as neutral spaces for communication. But this idea is becoming harder to accept. The case of X (formerly Twitter) under Elon Musk shows how platform decisions can reshape online environments rather than simply host them. A recent study found that hate speech on X increased by around 50% after changes to moderation policies, along with a 70% rise in engagement with such content, meaning that harmful posts were not only more common but also more visible (Ng et al., 2024).
At the same time, there was no meaningful reduction in inauthentic or automated accounts, despite claims that loosening moderation would improve the platform (Ng et al., 2024).
This shows that platforms are not as neutral as they claim to be. Digital systems are shaped by power, resources, and human decisions, rather than operating independently (Crawford, 2021).
Platform design also plays a key role in shaping user behaviour, meaning that harm is not simply the result of individual users but emerges from the system itself (Moore & Tambini, 2018).
This article looks at why platform governance is not just about removing harmful content after it appears.Instead, it is about shaping the conditions that allow such content to be produced, spread, and normalised.
The platform is not neutral.
（https://www.brookings.edu/articles/social-media-companies-and-common-carrier-status-a-primer/）
Platforms are often described as neutral spaces where users freely express themselves. But this idea ignores how platforms are actually designed and managed. Platforms don’t just host content—they shape how it is created, shared, and seen.
AI and platform technologies are built on specific social, economic, and political conditions, rather than operating as independent or objective systems (Crawford, 2021). This means that what appears online is not just the result of user behaviour, but also of underlying platform structures.
Platform design plays a crucial role in this process. Features such as recommendation systems, visibility rankings, and interaction mechanisms all influence what content gains attention. Rather than passively reflecting user preferences, these systems actively organise and prioritise information.
As a result, certain types of content—especially those that generate strong reactions—are more likely to spread widely. This helps explain why harmful content can become highly visible even without being explicitly promoted.
The case of X illustrates this clearly. Following changes to moderation policies, hate speech not only increased but also became more visible and more widely engaged with (Ng et al., 2024). This suggests that platform governance decisions directly affect what kinds of content thrive. Instead of being neutral intermediaries, platforms function as active environments that shape online discourse.
How algorithms amplify hate content.
If platforms aren’t neutral, the next question is simple: how do they shape what we see? A key part of the answer lies in algorithms. Recommendation systems and ranking mechanisms are designed to prioritise content that generates engagement, such as likes, shares, and comments. In reality, this often means that more extreme or emotional content spreads faster. Rather than simply reflecting user interests, these systems can amplify harmful content because it attracts attention.
This dynamic has been observed across multiple platforms. Online environments can develop what has been described as “toxic technocultures,” where platform features and user behaviour interact to normalise harmful content (Massanari, 2017). For example, systems such as upvoting, reposting, and algorithmic ranking can make certain viewpoints more visible, especially when they provoke strong reactions. Similarly, platform infrastructures can contribute to the spread of racism and discrimination by shaping how content is circulated and prioritised (Matamoros-Fernández, 2017).
The case of X reflects this pattern. The increase in engagement with hate speech suggests that such content was not only present, but actively circulating through the platform’s visibility systems (Ng et al., 2024). This supports the idea that algorithms do not just distribute content—they help determine which content becomes influential. As a result, online harm is not only a matter of individual behaviour, but also of how platforms are designed to reward and amplify certain types of content.
Platform responsibility and governance failure.
(Donald Trump stated he was ‘very happy that Twitter is now in sane hands’ after Elon Musk took over Twitter.)
If platform design and algorithms shape online spaces, then it’s hard to ignore the question of responsibility. Harmful content cannot be explained solely as the result of individual users, but must also be understood in relation to platform governance. The concept of a “duty of care” suggests that platforms have a responsibility to design systems that minimise harm rather than simply react to it after the fact (Moore & Tambini, 2018). This shifts the focus from content removal to system design, including how content is recommended, moderated, and circulated.
In reality, though, this responsibility is often not taken seriously. Content moderation is frequently inconsistent, limited, or shaped by commercial priorities rather than user safety. For example, research on platform governance in different regions has shown that moderation systems often fail to account for local context, making it difficult to effectively identify and address harmful content (Sinpeng et al., 2021). In some cases, responsibility is effectively pushed onto users, who are expected to report harmful content or manage their own online safety.
The case of X highlights these challenges. Despite a clear increase in hate speech and harmful content, there has been little evidence of effective intervention or system-level change (Ng et al., 2024). This suggests a gap between platform responsibility and platform practice. Rather than preventing harm through design, platforms may allow harmful content to spread while maintaining the appearance of neutrality or free expression. As a result, governance becomes reactive, fragmented, and often ineffective.
Power, business, and structural reasons.
So why does harmful content keep spreading? To answer that, we need to look beyond individual platforms and examine the broader structures they operate within. Digital platforms are not just communication tools, but powerful socio-technical systems shaped by economic interests, political forces, and institutional arrangements (Flew, 2021). This means that decisions about content moderation are not only technical choices, but also reflect deeper priorities such as growth, engagement, and profitability.
From this perspective, harmful content isn’t just a failure—it’s expected, but a predictable outcome of platform design. As Crawford (2021) argues, digital systems are built on forms of extraction, including user data and attention, which are central to platform business models. Content that generates strong emotional reactions—whether positive or negative—is more likely to drive engagement, making it valuable within these systems. As a result, there is an inherent tension between reducing harm and maintaining high levels of user activity.
The case of X reflects this structural dynamic. The increase in both hate speech and engagement suggests that harmful content may align with the platform’s underlying incentives rather than contradict them (Ng et al., 2024). This challenges the idea that platforms simply fail to manage harm. Instead, it suggests that the production and circulation of harmful content may be closely tied to how platforms are designed to operate.
Conclusion
The case of X makes it clear that online harm cannot be understood simply as a problem of individual behaviour or isolated content. Instead, it reflects deeper issues in how platforms are designed and governed. As this article has shown, harmful content is shaped by platform structures, amplified by algorithmic systems, and often left unaddressed due to weak or inconsistent governance. These dynamics challenge the common assumption that platforms are neutral spaces for communication.
Rather than treating harmful content as something that appears independently of platform systems, it is more accurate to see it as a product of those systems. Platform governance decisions—whether through moderation policies, algorithm design, or broader business models—play a central role in determining what kinds of content are produced and circulated online.
This leads to a bigger question about the future of digital platforms. If harmful content is built into the way platforms operate, can it be effectively reduced without fundamentally changing how these systems are designed? So fixing online harm may require more than just better moderation, but a rethinking of platform governance itself.
Reference
1.Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
2.Flew, T. (2021). Regulating platforms. Polity Press.
3.Massanari, A. (2017). #Gamergate and the fappening: How Reddit’s algorithm, governance, and culture support toxic technocultures. New Media & Society, 19(3), 329–346.
4.Matamoros-Fernández, A. (2017). Platformed racism: The mediation and circulation of an Australian race-based controversy on Twitter, Facebook and YouTube. Information, Communication & Society, 20(6), 930–946.
5.Moore, M., & Tambini, D. (2018). Digital dominance: The power of Google, Amazon, Facebook, and Apple. Oxford University Press.
6.Ng, L. H. X., Hui, P. M., & Lu, L. (2024). Hate speech and engagement on X (formerly Twitter) following changes to content moderation. PLOS ONE, 19(1), e0313293. https://doi.org/10.1371/journal.pone.0313293
7.Sinpeng, A., Martin, F., Gelber, K., & Shields, K. (2021). Facebook: Regulating hate speech in the Asia Pacific. Media International Australia, 178(1), 24–37.
1 Comment