Misleading Information and Clickbait

Much of the most consequential misinformation is not an outright lie. Misleading headlines and clickbait are hard for people to spot and common in both mainstream and unreliable outlets. Drawing on journalism and information studies, we built a taxonomy of the tactics they use and BaitBuster, a tool that flags misleading posts while people browse social media.

In BaitBuster 2.0, an NSF-funded collaboration, we turned to video. We curated a dataset of misleading video headlines with a multi-layered question-answering annotation strategy and built multimodal models that identify misleading videos with 88% accuracy. Current work tests how corrected headlines affect credibility and engagement, and whether large language models can identify and explain misleading tactics.