March 1, 2022

Timely and detailed information is critical during a disaster. With mainstreaming of social-media-based civic voluntarism during disasters, disaster managers often find themselves in a situation where they get overwhelmed by information streaming in through various channels. Real-time Emergency Assessment, Coordination, & Triage (REACT) AI is a research project aimed to automate the triage of this incoming image and text information and help disaster managers focus on areas that need the most attention. By working directly with Emergency Operations Centers in Japanese cities, this project closely integrates non-government and community-led organizations in disaster response, strengthening long-term resilience planning.

We are also conducting workshops with domain experts and disaster responders to understand the human factors affecting AI outputs and improve labeling criteria for training datasets. These workshops aimed to create tasks and labels consistent with the city’s protocol, are more objective for labelers, and are structured to help make operational decisions based on REACT output. In addition to piloting this model in real-life disaster drills, resulting datasets and underlying methodology will be open-sourced, helping drive the real-world impact of this project.

The project is supported by Google.org and undertaken in partnership with the University of Tsukuba, and Mercy Corps.