From physical space to cyberspace: Recessive gender biases in social media mirror the real world

In recent decades, our society has witnessed significant advancements in Information and Communication Technologies (ICT), leading to the creation of cyberspace, also known as virtual space. Within this digital landscape, social media has emerged as a crucial component of cyberspace, with the ability to both reflect and profoundly influence our physical world. But while positive content thrives, some negative contents are also taken into cyberspace from our physical world. Previous studies have discovered the existence of comments with strong gender biases in social media and proposed various efficient detectors to filter out biased information. However, the gender biases hidden behind the words, called recessive gender biases, receive less attention while being more widespread in social media. In this research, we propose 3-View Evaluation Framework (3VEF) to evaluate recessive gender biases embedded in social media using word embeddings. Taking Twitter data in Hong Kong as a case study, we apply the 3VEF framework to demonstrate the temporal trend and spatial distribution of recessive gender biases in social media in Hong Kong, and then discuss its relationship with sexism in physical space. The results reveal recessive gender biases in cyberspace mirror those in physical space with high consistency. Moreover, by integrating the spatial heterogeneity of recessive gender biases in social media with public statistical data in Hong Kong, we find that the factors affecting gender equality in physical space also play important roles in cyberspace. We believe our study contributes to understanding the recessive gender biases in cyberspace and exploring to avoid their further influence in physical space.