Overview of eRisk 2025: Early Risk Prediction on the Internet

Javier Parapar, Anxo Pérez, Xi Wang, Fabio Crestani

Abstract: This paper presents an overview of eRisk 2025, the ninth edition of the CLEF lab on early risk detection. Since its foundation, eRisk has served as a framework for evaluating methodologies, effectiveness metrics, and challenges related to the early identification of personal risks, particularly within the health and safety domains. The 2025 edition marks an important evolution of the lab, expanding its focus toward tasks that require deeper contextual and conversational understanding. In addition to continuing the depression symptom sentence ranking task, eRisk 2025 introduces two novel tasks: the second task is based on contextualized early detection of depression from full conversational threads, and a pilot task that explores the use of fine-tuned conversational agents for detecting signs of depression through interactions. These additions aim to open new research avenues and bring the evaluation setting closer to conversational scenarios.

External IDs:dblp:conf/clef/ParaparPWC25