CIRI research employs social media to enhance disaster response
During Super Storm Sandy, Twitter users sent more than 20 million tweets related to the massive storm (2013 National Preparedness report). This level of engagement with the platform could make Twitter and other social media platforms goldmines for disaster relief groups as they seek to improve internal processes and response times.
Agencies such as the Federal Emergency Management Agency (FEMA) and the US Coast Guard are increasingly looking to technology to help them achieve their mission of more quickly and safely distributing humanitarian aid, conducting rescue operations, and delivering other disaster response support. To support these response agencies, CIRI researcher Jana Diesner is incorporating artificial intelligence (AI) and data mining into the field of humanitarian assistance and disaster response (HADR).
Diesner, an associate professor of information sciences at the University of Illinois, is using AI and natural language processing techniques to analyze disaster-related data collected from social networks. She will then compare the data against national guidelines, such as DHS’ National Response Framework, to identify opportunities for improvement of those guidelines.
According to a FEMA spokesperson, “This research can help emergency managers save lives and make the best use of resources during highly stressful and chaotic events.”
This research takes into account three elements of disaster relief: the people affected, the disaster itself, and responding agencies such as FEMA and the United States Coast Guard.
Jana’s research is based on “ground truths.” These ground truths measure the results of her text mining against the actions of FEMA and the US Coast Guard. For instance, during a disaster if someone in an affected neighborhood posted to Facebook saying their electricity was out, how quickly was the electricity restored? Was other aid administered in this area? Using data mining to determine need could give an agency like FEMA insights into how best to respond to a disaster.
This project piggy backs off of Diesner’s earlier work in mining text data after the Haitian Earthquake in 2010. Initially, Diesner studied social media posts for clues on how best to distribute humanitarian aid. These posts and messages, mostly moving between Haiti and Florida, gave insights as to where aid like water, medicine, and housing was the most needed in Haiti.
Diesner’s work could help FEMA proactively improve its policies. If social media posts give a more reliable estimation of damage done and aid required, FEMA could begin to incorporate data mining into their emergency response tactics and be able to quickly and more effectively respond to a major disaster.