Original title: Beyond Truth Discovery: A Two-Stage Framework to Assess the Severity of False Claim during Disasters
Ruichen Yao, Tejna Dasari, Gulshat Baispay, Aizhan Zaurbek +4 · August 2026
During disasters, false rumors spread fast on social media. They can harm rescue operations and confuse the public. Most tools only flag whether a whole post is false. They do not pinpoint the exact lies inside or show how dangerous they are. The researchers built a two-stage system to fix this. First, a tool pulls specific false statements out of posts with text, images, videos, and links, then checks them against real evidence. Second, the system measures how severe each lie is by looking at two things: how believable it seems, and how much harm it could cause if people trust it. The team built a benchmark from real Reddit posts about hurricanes and wildfires. They discovered that older machine learning models struggle to match human views on severity. However, large language models match human thinking well, especially when guided with clear human examples.
Imagine you are a doctor sorting patients in an emergency room. Usually you just decide who is sick and who is not. But some patients have mild symptoms that could still be dangerous, while others look serious but are actually harmless. In this paper, the system treats false claims like those patients: it does not just label a claim as true or false, but scores how likely it is to be believed and how harmful it could be, so responders know which misinformation to tackle first.
During a disaster, not all false claims are equally dangerous. This framework helps emergency responders prioritize which misinformation to counter first, by scoring claims on how believable they are and how much harm they could cause. Focusing on the most severe false claims can improve public safety and trust.
The benchmark is built from Reddit posts about hurricanes and wildfires, so results may not generalize to other platforms, disaster types, or languages. The severity dimensions rely on human annotators with a shared rubric, which may not capture all cultural or contextual differences. The study does not test real-time deployment or scalability to millions of posts.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Beyond Truth Discovery: A Two-Stage Framework to Assess the Severity of False Claim during Disasters — Ruichen Yao, Tejna Dasari, Gulshat Baispay et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.