"Shut Up and Let Me Enjoy My Otome": Understanding and Measuring the Toxicity in Otome Game Communities
This paper introduces OtomeSCAN, a framework that analyzes 620,045 posts to measure and characterize pervasive toxicity in otome game communities, revealing significant platform disparities between Weibo and Reddit, the rapid escalation of toxicity during real-world conflicts, and the presence of coordinated attacks targeting game developers.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the vast landscape of digital entertainment, online communities have become the primary gathering places for millions of players. Within these spaces, a phenomenon known as toxicity has long been a subject of study for researchers. This term describes rude, disrespectful, or unreasonable comments that are likely to drive people away from a conversation. While much attention has been paid to hostility in general gaming groups, particularly those involving mixed genders, a specific and rapidly growing corner of the gaming world has remained largely unexamined. These are communities built around "otome" games, a genre of romance simulation where players step into the role of a female protagonist to build relationships with male characters. These games have evolved from niche hobbies into a global industry worth billions of dollars, attracting hundreds of millions of players. Yet, the social dynamics within these female-dominated spaces, and the specific nature of the abuse they face, have not been systematically measured until now. Understanding how toxicity manifests in these unique environments is crucial, not only for the players who seek safe spaces but also for the platforms that host them and the developers who create the games.
A team of researchers set out to fill this gap by conducting the first large-scale investigation into the toxicity found within otome game communities. They focused their study on two major social media platforms: Weibo, a Chinese platform similar to a mix of Twitter and Facebook, and Reddit, an English-language forum network. Over a period spanning eighteen months, from early 2024 to mid-2025, the researchers gathered and analyzed more than 620,000 posts. To make sense of this massive amount of text, they first created a specialized tool called OtomeSCAN. This framework allowed them to collect the data, clean it, and then use advanced artificial intelligence models to identify which posts were toxic. Because existing tools designed for general internet use often fail to understand the specific slang and cultural nuances of otome fans, the researchers trained their own models using a carefully hand-labeled dataset of over 4,300 posts. This human-verified data served as the ground truth, allowing the team to teach their AI to recognize insults, threats, and harassment specific to this community.
The results revealed a stark and surprising divide between the two platforms. On Weibo, the toxicity rate in otome communities was alarmingly high, with more than one in five posts containing harmful content. In contrast, the same communities on Reddit showed a toxicity rate of less than four percent, a figure nearly identical to the general gaming discussions found on that site. This suggests that the environment of the platform itself plays a massive role in how conflict is expressed. On the Chinese platform, the abuse was frequently directed at the players themselves, often attacking their identity or intelligence. On Reddit, however, the hostility was more likely to be aimed at the fictional characters within the games or the developers who made them, rather than the fans. This difference highlights that toxicity is not a uniform problem; it changes shape depending on the cultural and technical rules of the space where it occurs.
The study also tracked how these toxic outbursts change over time, finding that they often surge in response to real-world events. When a game released a new update, when a celebrity made a mocking comment about the players, or when a fan group felt their rights were violated, the rate of toxic posts would spike dramatically. In one instance, following an external attack on the community, the proportion of toxic posts on Weibo jumped from roughly twenty-two percent to over thirty-seven percent in just three days. The researchers observed that these spikes were not random; they followed a pattern where internal conflicts or external pressures would cause the community to turn inward, often targeting the game developers or the players themselves. The study identified nearly two hundred instances where groups of users appeared to coordinate their attacks, often using similar language or templates to harass developers, though the researchers noted that these were potential patterns of coordination rather than proof of a single organized conspiracy.
Beyond the numbers, the researchers looked closely at the language used in these toxic posts to understand how players tried to evade detection. On Weibo, nearly forty percent of the unique toxic phrases used clever tricks to hide their meaning, such as using slang, abbreviations, or substituting characters with homophones that sound the same but look different. This allowed the abusive content to slip past standard filters. On Reddit, the language was much more direct, with fewer attempts to disguise the insults. The study concluded that while otome games have built a massive and profitable industry, the communities surrounding them face unique challenges that general gaming research has overlooked. The findings suggest that a one-size-fits-all approach to moderating online spaces will not work; instead, platforms and developers need to understand the specific cultural and linguistic dynamics of these communities to build healthier environments for their players.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.