Soft Moderation
Soft moderation aims to mitigate the spread of misinformation online by subtly flagging potentially misleading content, rather than outright removal. Current research focuses on improving the accuracy and context-awareness of automated soft moderation systems, employing techniques like perceptual hashing for image analysis and contrastive textual deviation for nuanced stance detection to reduce false positives. These advancements are crucial for building more effective and trustworthy online platforms by providing more precise warnings and avoiding the unintended consequences of overly broad or inaccurate moderation.
Papers
PIXELMOD: Improving Soft Moderation of Visual Misleading Information on Twitter
Pujan Paudel, Chen Ling, Jeremy Blackburn, Gianluca Stringhini
Enabling Contextual Soft Moderation on Social Media through Contrastive Textual Deviation
Pujan Paudel, Mohammad Hammas Saeed, Rebecca Auger, Chris Wells, Gianluca Stringhini