[Paper Review] Hate Raids on Twitch: Echoes of the Past, New Modalities, and Implications for Platform Governance

Catherine Y. Han, Joseph Seering+3 more|arXiv (Cornell University)|Jan 10, 2023

Hate Speech and Cyberbullying Detection4 citations

TL;DR

This study investigates hate raids on Twitch—coordinated, identity-based attacks overwhelming streamers' chats with racist and antisemitic messages—using large-scale data and interviews with marginalized streamers and tool developers. It reveals that while most raids are hate-driven and target LGBTQ+ and Black streamers, a subset resembles subcultural trolling, prompting community-led defense strategies that highlight the need for platform-community collaboration in governance.

ABSTRACT

In the summer of 2021, users on the livestreaming platform Twitch were targeted by a wave of "hate raids," a form of attack that overwhelms a streamer's chatroom with hateful messages, often through the use of bots and automation. Using a mixed-methods approach, we combine a quantitative measurement of attacks across the platform with interviews of streamers and third-party bot developers. We present evidence that confirms that some hate raids were highly-targeted, hate-driven attacks, but we also observe another mode of hate raid similar to networked harassment and specific forms of subcultural trolling. We show that the streamers who self-identify as LGBTQ+ and/or Black were disproportionately targeted and that hate raid messages were most commonly rooted in anti-Black racism and antisemitism. We also document how these attacks elicited rapid community responses in both bolstering reactive moderation and developing proactive mitigations for future attacks. We conclude by discussing how platforms can better prepare for attacks and protect at-risk communities while considering the division of labor between community moderators, tool-builders, and platforms.

Motivation & Objective

• To understand the technical and social characteristics of hate raids on Twitch, particularly their targeting of marginalized identities.
• To examine the lived experiences of LGBTQ+ and Black streamers during hate raids and how these attacks impact their safety and sense of community.
• To analyze the responses of streamers, community moderators, and third-party tool developers to hate raids, including both reactive and proactive mitigation strategies.
• To explore the division of labor between platforms, communities, and power users in responding to coordinated harassment and to recommend structural improvements in platform governance.
• To advocate for intersectional, user-centered design that prioritizes the needs of the most vulnerable users in platform safety mechanisms.

Proposed method

• Conducted a large-scale quantitative analysis of chat data from 9,664 popular Twitch channels to detect and characterize hate raids based on message content and timing patterns.
• Used community-aggregated lists of known malicious bots to identify and filter hate raid activity, enabling detection of coordinated attacks.
• Conducted in-depth, semi-structured interviews with seven LGBTQ+ and/or Black Twitch streamers to understand the emotional, social, and structural impacts of hate raids.
• Interviewed two third-party bot developers to examine the technical design and deployment of community-built moderation tools in response to raids.
• Combined qualitative insights with quantitative data to identify dual modes of hate raids: hate-driven attacks and subcultural trolling behavior.
• Analyzed Twitch streamer tags to assess whether attackers used self-identified community tags to target marginalized streamers.

Experimental results

Research questions

• RQ1 RQ1: What are the fundamental characteristics of hate raids—how are they orchestrated, and who do they target?
• RQ2 RQ2: How do hate raids affect members of marginalized communities, particularly Black and LGBTQ+ streamers?
• RQ3 RQ3: How did different stakeholders—streamers, community moderators, and tool developers—respond to hate raids, and what strategies emerged?
• RQ4 RQ4: How can platforms better integrate community knowledge and tools into their governance models to prevent future attacks?
• RQ5 RQ5: What role do structural inequalities and platform design play in enabling or mitigating hate raids?

Key findings

• 98% of hate raid messages on Twitch were identity-based attacks, with anti-Black racism and antisemitism being the most prevalent ideologies.
• Despite the hate-driven nature of most raids, attackers often targeted streams indiscriminately, though evidence suggests they used streamer tags (e.g., Black, LGBTQ+) to identify marginalized communities.
• LGBTQ+ and Black streamers reported that hate raids were highly targeted, persistent, and part of broader harassment campaigns, sometimes extending to offline threats like swatting.
• The community responded with both reactive moderation (e.g., muting, banning) and proactive tool development, including third-party bots to filter hate content in real time.
• Power users and community tool developers, such as Sery, played a critical role in mitigating raids, suggesting that platforms should consult and compensate such contributors.
• The study highlights a governance gap: platforms like Twitch often fail to act on community-identified threats, leading to reliance on volunteer-led defenses that are unsustainable without institutional support.