Generalizing to Unseen Disaster Events: A Causal View

TL;DR

This paper introduces a causal approach to mitigate biases in social media disaster data, improving the generalization of classifiers to unseen disaster events and outperforming existing baselines.

Contribution

It proposes a novel causal bias mitigation method tailored for disaster event classification, enhancing model generalization to new, unseen events.

Findings

Improves F1 score by up to +1.9% over baselines

Significantly enhances classifier performance across three disaster tasks

Reduces event- and domain-related biases effectively

Abstract

Due to the rapid growth of social media platforms, these tools have become essential for monitoring information during ongoing disaster events. However, extracting valuable insights requires real-time processing of vast amounts of data. A major challenge in existing systems is their exposure to event-related biases, which negatively affects their ability to generalize to emerging events. While recent advancements in debiasing and causal learning offer promising solutions, they remain underexplored in the disaster event domain. In this work, we approach bias mitigation through a causal lens and propose a method to reduce event- and domain-related biases, enhancing generalization to future events. Our approach outperforms multiple baselines by up to +1.9% F1 and significantly improves a PLM-based classifier across three disaster classification tasks.