This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: Online spaces, and social media in particular, are forums for vigorous political debate. They provide a space for members of the public to express political opinions, organize and sometimes even interact directly with politicians.
While they clearly have democratizing potential, these platforms have also become fertile ground for hate speech, toxic interactions and online violence. By analyzing more than 3.2 million responses on X (formerly Twitter) to 88 candidates for the 2024 European Parliament elections, our team at the European Commission's Joint Research Centre found significant differences in how women and men in politics are treated online. Applying a toxicity detection algorithm called Perspective API, we studied both the volume and types of "toxic" replies received by different candidates.
The algorithm's toxicity score ranges from zero to one. It estimates the likelihood that a reader would perceive text as toxic and defines toxicity as "a rude, disrespectful, or unreasonable comment that is likely to make you leave a discussion." Almost 80% of the candidates studied, who were from France, Ireland, Italy, Portugal and Spain, received toxic replies, but the raw numbers showed a real and measurable gender gap: Women candidates received more toxic replies in absolute terms than their male counterparts. However, accounting for differences in the overall number of replies for each post, the gender difference in the proportion of toxic replies was no longer statistically significant.
Women posted less than men and yet were addressed more. In other words, their accounts attracted more attention overall, toxic and benign alike. To conduct our analysis, we gathered posts mentioning each candidate, then narrowed the data to focus on replies to the candidates' posts.
The toxicity detection algorithm scored each post for toxicity, flagging anything above a given threshold for closer inspection. Once the algorithm had done its work, we turned to what the messages actually said, manually coding nearly 2,700 of the most clearly toxic replies. These were messages where the automated tool registered the highest probability of toxicity, meaning nine out of 10 human raters would agree they were toxic.
The content of the posts clearly showed that women and men face vastly different attitudes once they enter the realm of politics. Of these clearly toxic replies, 71% were directed at women candidates. The content was strikingly different from what male politicians faced.
Extract — continue reading at the source.
