Shirtless and Dangerous: Quantifying Linguistic Signals of Gender Bias in an Online Fiction Writing Community

Imagine a princess asleep in a castle, waiting for her prince to slay the\ndragon and rescue her. Tales like the famous Sleeping Beauty clearly divide up\ngender roles. But what about more modern stories, borne of a generation\nincreasingly aware of social constructs like sexism and racism? Do these\nstories tend to reinforce gender stereotypes, or counter them? In this paper,\nwe present a technique that combines natural language processing with a\ncrowdsourced lexicon of stereotypes to capture gender biases in fiction. We\napply this technique across 1.8 billion words of fiction from the Wattpad\nonline writing community, investigating gender representation in stories, how\nmale and female characters behave and are described, and how authors' use of\ngender stereotypes is associated with the community's ratings. We find that\nmale over-representation and traditional gender stereotypes (e.g., dominant men\nand submissive women) are common throughout nearly every genre in our corpus.\nHowever, only some of these stereotypes, like sexual or violent men, are\nassociated with highly rated stories. Finally, despite women often being the\ntarget of negative stereotypes, female authors are equally likely to write such\nstereotypes as men.\n