Parsing tweets to strengthen community disaster resilience
Extreme weather and climate events cause significant loss of life and property, as well as substantial social and economic loss. Compounding the problem, population growth in vulnerable areas, resource constraints and climate extremes make it difficult to rapidly recover interrupted services — especially in the context of climate change, as we face a future in which disasters will become more severe or more frequent. And the COVID-19 pandemic has added new needs.
A huge issue facing towns and cities in the long term is how to integrate disaster preparation, response and recovery into the fabric of their communities to enable better outcomes.
To improve disaster response and resilience, our team in the Laboratory for Advancing Sustainable Critical Infrastructure is developing algorithms, based on extensive studies in hazard science, that examine how people react to disasters on social media. Twitter is among the most popular social media platforms, with some 64 million monthly active users in the United States. Our algorithms search through tweets for messages that relate to the major components of a resilient community — namely, its social, ecological, economic and institutional, and individual health. These parts constitute the resilience fingerprint — the underlying elements of what makes a community hardy enough to cope with disasters.
For example, social health includes messages that involve non-institutional support systems, like neighbors, volunteers and humanitarian aid, while the ecological aspect relates to natural systems, such as streams, beaches and the coast. The economic side refers to financial components, as well as elements including business operations and currency, and institutions relates to government and service bodies, such as police departments, hospitals and the Federal Emergency Management Agency (FEMA).
Our innovation goes beyond off-the-shelf machine learning and AI tools to create the custom-built algorithms that crunch this data. We analyzed almost 120 million tweets spanning 14 major events — from hurricanes to earthquakes, political events and public health crises — and found significant commonalities in how people respond to each kind of occurrence. For instance, the tweets from the hurricanes exhibited lots of discussion around specific environmental and ecological impacts that separated hurricanes clearly from, say, earthquakes. The algorithm also can be used to assess community resilience during such public health crises as the COVID-19 pandemic.
A critical component in disaster response and recovery is how resourceful individuals within a community are and how well they work together to help one another. Social media allows a glimpse of what members of the community are thinking and doing during disasters, giving us a better sense of how the community as a whole is responding to the crisis at hand.
What is even more vital is that we can see these thoughts and actions in near real time to reach a higher level of disaster response. The goal is to make sure that what decision makers see as important for building a resilient community is also what the community values. This helps these decision and policy makers better prepare and plan for mitigation and recovery efforts and make more intelligent, targeted investments in a future resilient grid.
Resilience means varied things, depending on whether you are talking to engineers building bridges or social workers aiding in community care. Unifying all of these perspectives is the concept of “bouncing back” — a resilient community is one in which all segments of the population work together to anticipate, adapt and respond to any kind of disruption or disaster.
Our goal is to understand, measure and harness how social, technological and environmental systems interact during disasters, in order to develop resilient communities that can face an uncertain future with confidence rooted in sound preparedness and the indomitable human spirit.
Roshi Nateghi, PhD
Assistant Professor of Industrial Engineering
College of Engineering, Purdue University
Laboratory for Advancing Sustainable Critical Infrastructure
School of Industrial Engineering
College of Engineering, Purdue University