Overview of eRisk 2024: Early Risk Prediction on the Internet (Extended Overview)

Abstract

[Abstract]: This paper presents eRisk 2024, the eighth edition of the CLEF conference’s lab dedicated to early risk detection. Since its inception, the lab has been at the forefront of developing and refining evaluation methodologies, effectiveness metrics, and processes for early risk detection across various domains. These early alerting models hold significant value, particularly in sectors focused on health and safety, where timely intervention can be crucial. eRisk 2024 featured three main tasks designed to push the boundaries of early risk detection techniques. The first task challenged participants to rank sentences based on their relevance to standardized depression symptoms, a crucial step in identifying early signs of depression from textual data. The second task focused on the early detection of anorexia indicators, aiming to develop models that can recognize the subtle cues of this eating disorder before it becomes critical. The third task was centered around estimating responses to an eating disorders questionnaire by analyzing users’ social media posts. Participants had to leverage the rich, real-world textual data available on social media to gauge potential mental health risks. Through these tasks, eRisk 2024 continues to advance the field of early risk detection, fostering innovations that could lead to significant improvements in public health interventions.