Researchers develop AI tool to improve crisis communication

During a crisis, emergency managers require swift, accurate data to make informed decisions, but the overwhelming and evolving influx of information during such events can make it challenging to distinguish facts from misinformation. Researchers at Texas A&M University are addressing this issue by creating an AI tool capable of sorting through the chaos.

Led by postdoctoral researcher Dr. Bo Li, along with Dr. Ali Mostafavi and Kai Yin from the Zachry Department of Civil and Environmental Engineering, the team employs a large language model to filter the deluge of data generated during disasters like hurricanes.

In a recent study published in Advanced Engineering Informatics, the researchers utilized a collection of human-annotated social media posts from various disaster incidents to train the model. The system is designed to classify crisis-related information into three categories: disaster-related, weather-related, and humanitarian information.

Moreover, it further classifies each post based on event type, informativeness, and the specific type of humanitarian assistance or rescue needed. By categorizing data in this manner, emergency personnel can more readily access the specifics pertinent to their requirements, such as updates on storm conditions, reports of damage, and requests for aid and rescue assistance.

Although the system is still in the development phase and has not been implemented during an active disaster, the researchers have ambitious plans to evolve the tool further. Li envisions the creation of an AI agent that emergency professionals can interact with during a disaster. This interactive AI system would aim to enhance situational awareness, moving beyond mere classification to actively support decision-making processes.

As disasters continue to occur more frequently and the volume of available information grows exponentially, the researchers believe that AI-based tools have the potential to significantly reduce the time spent sifting through reports, allowing emergency response agencies to swiftly focus on making decisions that directly benefit affected communities.