Suicidal Ideation Detection in Incel Forums

Loading...
Thumbnail Image
Date
Authors
Van Herk, Henry
Keyword
dark web forums , incel characteristics , suicide detection , natural language analytics
Abstract
Online communities allow like-minded individuals from across the globe to connect effortlessly. Many communities enforce guidelines to ensure safe user experiences, while others, such as the incel community, allow free speech and promote having no censorship. While many discussions in incel forums deal with their shared hate for women, there are numerous posts about individuals’ suicidal thoughts and plans. On top of suicide being a severe public health problem, the emergence of lone wolf terrorist attacks perpetrated by individuals with incel ideologies makes incel forums an important online channel to monitor. Analyzing online posts in these forums dealing with suicidal ideation can help identify individuals who may be likely to die by suicide or carry out such an attack, which can be helpful to organizations that wish to prevent suicide attempts and possible lone-wolf terrorism. While suicidal ideation detection in online user content is a well-researched problem, not much research involves incel forums. My work involves research in suicidal ideology, incel terminology, and existing suicidal ideation detection methods outside of incel forums. Multiple classical supervised learners, as well as deep learners, were trained on an incel forum gathered from the dark web. The resulting best-performing suicide intent detection model achieved an F1 score of 97.5%. Once calculated, these predictions could prioritize posts and their authors for investigation.
External DOI