CU Boulder pioneers culturally sensitive AI solutions for disasters

Caption: Participants of the Harnessing Artificial Intelligence (AI) for Disaster Management: Bridging Research, Practice, and Community Engagement workshop held on the CU Boulder campus.

In disaster and emergency management, rescue teams rely on simple tools, like drones and mobile devices connected to the internet, to take photos of damaged buildings and document the resulting destruction. First responders use this information to prioritize where help is needed immediately and in the long-term.

At the same time, many emergency response agencies have been slow to embrace advanced tools like artificial intelligence (AI) due to concerns of complexity, training needs, initial costs and the potential impact on their work, said Professor Amir H. Behzadan, of CU Boulder's Department of Civil, Environmental and Architectural Engineering and a faculty research fellow of the Natural Hazards Center at the Institute of Behavioral Science.

"This hesitation, along with a historical lack of diversity in the workforce, has contributed to delayed and unfair treatment of communities hit hardest by disasters," Behzadan said. "Lower-income disaster survivor homeowners are not only less likely to receive assistance compared to wealthier counterparts, but also tend to receive substantially less aid."

On April 19, Behzadan hosted a workshop titled "AI4DM - Harnessing Artificial Intelligence (AI) for Disaster Management: Bridging Research, Practice, and Community Engagement," with participants from academia, disaster management practitioners and disaster nonprofits. Attendees included representatives from CU Boulder, Texas A&M University, University of Florida, University of North Texas, University of Texas at Arlington, and University of Albany along with the Colorado Office of Emergency Management.

The workshop was supported by Behzadan's NSF grant from the Future of Work: Human-Technology Frontier program, which aims to raise awareness of AI and related technologies in disaster management. Participants were introduced to human-centered AI applications in disaster management and encouraged to work toward ways to adopt AI-informed solutions.

"AI can be applied in disaster management for various purposes, such as generating instant disaster data visualizations; providing interview training and reskilling/upskilling for employees; assessing and mapping disaster damage; and enhancing disaster risk messaging and communication strategies," Behzaden said.

Human-centered AI focuses on designing AI systems that prioritize human needs, values, expectations and experiences, aiming for beneficial, equitable, safe and ethical outcomes. In disaster and emergency management, this includes increased representation of data from marginalized groups, using inclusive communication strategies that resonate with at-risk communities, considering linguistic and cultural sensitivities and removing personal biases when assessing community needs during disasters.

"Human-centered AI leads to more effective disaster response and recovery practices by promoting collaboration, understanding and inclusivity among team members, survivors and stakeholders," Behzadan said. "It helps create equitable solutions that address the needs of all affected communities, regardless of their socioeconomic and cultural backgrounds."

Behzaden and his team plan to pursue additional funding to sustain these efforts. While leveraging their existing partnerships in Texas and Florida, they are also forging new relationships with local communities as well as disaster and emergency management organizations in Colorado who will ultimately be the beneficiaries of these technologies.

"The success of any technology hinges on its adoption by the intended end users," Behzadan said. "Understanding user needs and preferences, especially when working with vulnerable groups, is crucial for meaningful adoption and realizing the full potential of technological advancements."