Skip to main content

Interdisciplinary Team Awarded $1.2M to Explore the Future of Risk Prediction in Fire Departments

The word “firefighter” likely brings to mind images of someone battling the flames of a burning building. But what about preparations leading up until that moment? What about the firefighter who visits a school to talk about fire safety or who visits the home of an elderly person to discuss fall prevention?

Yunan Chen
Yunan Chen

“The majority of their time is not spent putting out fires,” says Yunan Chen, associate professor of informatics in UC Irvine’s Donald Bren School of Information and Computer Sciences (ICS). “They’re actually doing more risk prevention work for the local community.”

Chen is part of an interdisciplinary team working to help fire departments better analyze data and apply prediction information technology (IT) for community risk reduction (CRR). The work is being supported through a $1.2 million National Science Foundation grant, “Understanding the Present and Designing the Future of Risk Prediction IT in Fire Departments.” Chen is collaborating on the NSF Human-Centered Computing grant with her colleague, Melissa Mazmanian, Chancellor’s Fellow and chair of the Informatics Department; Mauricio Mejía and Kathleen Pine of Arizona State University; and Myeong Lee of George Mason University.

The three-year project began in September 2022 with ethnographic research of fire departments and their daily work practices. This will lay the foundation for the future design and development of new data-driven risk prediction principles and management (DDRPM) tools. Such tools can aid fire departments in anticipating and managing a variety of community risks, including those related to medical, fire and safety emergencies.

Community Risk Reduction
As outlined in the grant abstract, CRR is a paradigm that seeks to use DDRPM tools to mitigate risks before they lead to emergencies. Analyzing data such as past weather and wind patterns or 911 calls could help fire departments predict future trends and better prepare for — and possible prevent — emergencies.

Yet the researchers stress in the grant that “understanding the intersection of sociotechnical work practices for handling risks and the design of data-driven computational tools requires fundamental human-centered computing research.” The team is thus taking a three-phased approach: documenting current practices in three fire departments through ethnographic research, codesigning speculative DDRPM prototypes with study participants, and evaluating the prototypes and their impact on visions of technologically supported community risk work for the fire service.

Ethnographic Research

Rachel Warren headshot
Rachel Warren

Mazmanian, Yunan and informatics Ph.D. student Rachel Warren have started with an empirical exploration of community-oriented risk work performed by fire personnel, as well as any associated data work. Working with fire departments in California, Arizona and Virginia, they are mapping out current communication and resource allocation processes and identifying the various information repositories and IT systems in use.

“Who’s collecting the data? Who’s analyzing the data? Who’s building the models to make the predictions?” asks Chen. “And what is the relationship between human agency and technological-mediated agency?” They are also examining collaborative processes, looking at how data sharing works with local hospitals, transportation and police departments, and elder care centers. These are all critical questions to ask as fire departments move to more data-driven practices, particularly given the problematic implementation of predictive technology in other areas.

MelissaMazmanian
Melissa Mazmanian

“There’s a lot of scholarship suggesting the problematic applications of predictive analytics and the ways in which the inevitable biases of these algorithms — usually trained with past data [with] systemic bias in it — perpetuate and also can intensify issues,” says Mazmanian. Concerns with predictive healthcare and policing applications, for example, have already been documented. However, this is still a fairly new practice for fire departments, which are only in the initial stages of exploring how data analytics will transform their work. “So, it’s a great time to intervene to better understand what’s being built,” says Mazmanian, “and to potentially have a design intervention.”

Looking to the Future
Once the team has a solid understanding of current practices, the goal is to continue working closely with stakeholders to codesign a software prototype that would improve information workflow and analytics, helping fire departments set priorities and allocate resources without overburdening users.

“We’re hoping that, in the process, people actually embrace this future risk prediction technology,” says Chen. “We’re going to use a practice called speculative design, where you speculate what the future could be to help the fire department understand the potential work practice transformation.” In addition to applying new software, this could involve hiring additional people and reconfiguring workflows, but it depends on the department size and funding. Not every department can create a data science team to curate, analyze and communicate the data. “There are fire departments in large cities and smaller countryside towns. Every place is different, so how do we account for that?” asks Chen. The team will address this and other challenges using their novel human-centered approach centered around participatory and speculative design.

“It’s a new project that we’re really excited to be getting off the ground,” says Mazmanian. “We see predictive analytics as holding this promise for better deployment of services.”

— Shani Murray