The field of online harm mitigation is moving towards a more nuanced understanding of the complex interactions between users, content, and context. Researchers are developing innovative methods to detect and analyze abusive language, censorship, and strategic antisocial behavior online. A key direction in this field is the incorporation of contextual information, such as conversational exchanges and user demographics, to improve the accuracy of detection models. Another important area of research is the study of disengagement from problematic online communities, which can provide valuable insights for designing interventions to counter harmful ideologies. Noteworthy papers in this area include:

• A study on incorporating target awareness in conversational abusive language detection, which found that leveraging context from parent tweets leads to substantial improvements in classification performance.
• Research on state and geopolitical censorship on Twitter, which developed a user-level binary classifier to predict whether an account is likely to be withheld.
• An analysis of exit stories from problematic online communities, which identified a range of factors contributing to disengagement and highlighted the need for moving beyond interventions that treat conspiracy theorizing solely as an information problem.
• A paper on identifying strategic antisocial behavior online, which developed a new tree-structured Transformer model to categorize replies based on their hierarchical conversation structures and found a strong correlation between the presence of attackers' interactions and chilling effects on journalists.