SIFI-H: Social Impact of False Information dur- ing Hurricanes

SIFI-H is built upon a large-scale, cross-platform dataset focused on the 2024 Atlantic Hurricane Season, one of the most active and costly seasons in recorded history, to ensure our dataset reflects timely and diverse online information. The sampled posts are then manually labeled by domain experts from our team, who assess each post’s potential societal impact based on their disciplinary expertise. The annotation process follows predefined guidelines that consider factors such as the potential to mislead, harm, and impact the public. Each post is assigned a social impact label on a binary-level scale (i.e., low, high) to reduce annotation ambiguity and improve inter-annotator agreement.