2026
EMNLP
EMNLP 2026
Adaptive Causal Coordination Detection for Online Harm Campaigns: A Framework Study with Semi-Supervised Labeling
Abstract
AbstractACCD detects coordinated online harm campaigns by combining adaptive causal temporal scoring, semi-supervised account classification, and stability validation. On a filtered Twitter IRA task, it improves account-level coordination screening over correlation and fixed-parameter CCM baselines, while reducing manual labeling and runtime.