This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: During a disaster, emergency managers need quick, reliable data to make decisions, but the vast and changing stream of information that emerges during a crisis can make it difficult to distinguish facts from noise. Researchers from Texas A&M University are working to change that through the development of a tool that uses artificial intelligence to sort through the confusion.
Led by postdoctoral researcher Dr. Ali Mostafavi and Kai Yin in the Zachry Department of Civil and Environmental Engineering, the work uses a large language model to filter the information that emerges during crises such as hurricanes, when emergency officials must make quick decisions while monitoring weather conditions, damage reports, public requests for assistance, search-and-rescue needs and more. In their recent study on crisis-related social media classification, published in Advanced Engineering Informatics, researchers used a benchmark of human-annotated social media posts from multiple disaster events.
The model was designed to categorize crisis-related information into three areas: disaster-related, weather-related and humanitarian information. The model goes even further into classifying each post by event type, informativeness and humanitarian information type. Li said organizing information in categories could allow emergency personnel to more easily locate details relevant to their needs, such as updates on storm conditions, reports of damage or requests for aid and rescue.
The system is under development and has not been deployed during an active disaster. "The team plans to build an AI agent that emergency professionals can talk to when disaster strikes," Li said. "Ultimately, we want to move from classification toward a more interactive AI system that can better support situational awareness during disasters." As disasters become more frequent and the amount of available information continues to grow, the researchers say AI-based tools could help response agencies reduce the time spent sorting reports and focus more quickly on decisions that support affected communities.
Kai Yin et al., CrisisSense-LLM: instruction fine-tuned large language model for multi-label social media text classification in disaster informatics, Advanced Engineering Informatics (2026). DOI: 10.1016/j.aei.2026.104836 BA art history, MA material culture. Former museum editor, paramedic, and transplant coordinator.
Editing for Science X since 2021. Full profile → Master's in physics with research experience. Long-time science news enthusiast.
Extract — continue reading at the source.
