Can AI tools help in an emergency? Research is guiding that question
As more industries consider using artificial intelligence (AI) for efficiency, public sector managers face questions. What tasks are appropriate to offload to AI and which ones are best left to human experts?
Approximately 1,400 miles of pipes, 6,600 storm structures, and hundreds of miles of manmade channels. This is the extensive Fairfax County stormwater system that Chase Suddith, a Virginia Tech alumnus and an emergency management specialist for the county, oversees as part of his job.
He’s responsible for managing snow, floods, and related weather events for a county of approximately 1.2 million people.
Could artificial intelligence (AI) make Suddith’s job easier? Right now, like many people, Suddith uses AI tools to help craft emails and create meeting agendas.
“It’s good at creating checklists and main summaries,” said Suddith, who earned a master’s degree in public administration from Virginia Tech in 2018. “It’s good at communicating and distilling information down.”
But when it comes to using AI in other work, some emergency managers like Suddith are cautious.
Suddith is one of 30 emergency managers in the greater Washington, D.C., metro area who is helping to inform a Virginia Tech research team’s ongoing analysis to determine the ways that AI tools can help crisis and emergency managers do their work better. The team is led by Shalini Misra, associate professor in the School of Public and International Affairs and director of the Public-Interest Technology Lab, based at the Virginia Tech Research Center — Arlington. The lab includes faculty from various disciplines who research the governance, design, and deployment of new, emerging technologies with an eye on human impact.
This project analyzing AI tools is particularly timely. As more industries consider using AI for efficiency, public sector managers face questions — what tasks are appropriate to offload to AI and which ones are best left to human experts?
Misra said her team decided to focus on emergency managers because of the important public roles they play. The research was funded by a Destination Areas 2.0 award from Virginia Tech with the goal of designing, implementing, and evaluating AI tools that serve the public interest along with the managers who use them.
“We are in an era of cascading crisis and disasters. You might say we are always in a state of crisis these days,” she said. “They are working in many levels of the government. They could be at the federal level but also work in the city and county level, and their work involves coordination across levels of government and other agencies to be able to respond to disasters. They have to be able to collaborate with schools, hospitals, and with emergency responders. We wanted to understand the potential impact of AI on a particular type of public manager who is very important to the cycle of problems that we are facing right now in an era of disasters and crisis.”
The project
The research team’s findings are published in the journal Government Information Quarterly, and they feature the results of surveys with emergency managers across the United States. The surveys compiled information about how those who work in an emergency management capacity view AI and whether and why they would be willing to use AI tools for certain managerial tasks.
In August, Misra and her team hosted emergency managers for a workshop at the Virginia Tech Research Center. The goals were to learn about the barriers and challenges to incorporating AI tools in their work; how AI tools could be designed for transparency, accountability, and justice; and to test a tool for hazard mitigation planning.
The team will host other workshops with this same group next year in Arlington.
Some emergency managers said they have used AI for planning tasks but not during an actual emergency, when response time is critical and the public needs quick and accurate communication and decision making.
“Their decisions affect lives, and they have to be transparent and accountable in the end,” Misra said. “So when they need it the most, AI tools need to be designed to be equitable, accurate, transparent, and accountable.”
Still for planning ahead for disasters, such as creating a hazard mitigation plan, AI could be most useful, the Virginia Tech team found.
Hazard Helper
The team is testing a specialized AI tool for hazard mitigation planning called Hazard Helper. It would allow emergency managers to create or update hazard mitigation plans ahead of a crisis and eventually incorporate other helpful information into them, such as census data and accessibility details. Some departments hire companies to create these kinds of plans, so an AI tool could help with efficiency and save money.
In the world of emergency management, there are “periods of planning and pre-work, and there’s the time of the response or preparing for something imminent to happen,” said Ben Katz, a member of the research team and associate professor in the Department of Human Development and Family Science. Katz’s work is focused on how AI can reduce cognitive load on those who use it.
A tool like Hazard Helper would be useful in taking large quantities of information and distilling it into one comprehensive hazard mitigation plan, Suddith said. It also could be helpful for emergency managers who may be a staff of one in a smaller government office.
As the team’s work continues next year, this project is a microcosm of its ultimate charge.
“Instead of thinking about justice, equity, bias, and human agency as an afterthought after these technologies are out in the world, we think about them at the design stage and before implementing them,” said Misra. “We think together with the professionals who are going to use them, and the community members who it will affect.”
