Using AI to Build Community-driven Urban Resilience

Examples of AI-generated images created by the Asia Resilient Cities project. Photo credit: Asia Resilient Cities

By Stephanie Wasserman and Syrga Kanatbek Kyzy

Artificial intelligence (AI) has made headlines in a range of fields for its potential to make both welcome and unwelcome changes. The former includes helping researchers collect and visualize data to increase understanding between policymakers and communities. Through the USAID-funded Asia Resilient Cities (ARC) project, we use AI-generated images to help us communicate strategies and solutions for sustainable urban growth in our partner cities.

In Rajkot, India, AI-generated images help us hold more effective community feedback sessions (CFSs). CFS are a key part of ARC’s systems thinking approach—we collaborate with a range of stakeholders to co-create practical, sustainable solutions for the city’s most pressing resilience and climate concerns, ensuring partnerships with residents and city governments to build community-driven solutions. Discussions with CFS groups help us refine our proposed solutions and incorporate new perspectives. In Rajkot, ARC held CFS with frontline workers from city government departments; women, including those in care-taking roles and working outside the home; students of diverse socioeconomic backgrounds; and migrant workers who reside in the city. Through these conversations, we learned about the city’s key climate challenges such as rising temperatures, water scarcity, and waterlogging.

Why did ARC Use AI-generated Images?

AI-powered images offered a way for ARC to effectively share proposed solutions to Rajkot’s key climate challenges. The visuals also allowed us to share the same information with different CFS groups, regardless of literacy level. For each challenge we created a set of AI-generated images depicting the issue and how the proposed solution could solve it. While the example images below have English dialogue bubbles, the versions used in the Rajkot CFS were in Gujarati, the primary language of most participants.

In speaking with city stakeholders, ARC learned that one of the climate-related challenges that Rajkot residents are most concerned about is rising temperatures. Image 1, the “issue,” shows the effect of urban heat islands—caused by limited green space and high concentration of buildings, roads, and sidewalks—on dismayed residents.

Image 2, the “solution,” portrays a range of city stakeholders discussing strategies such as green spaces. The city in the background looks cooler, and the dialogue indicates that residents are collaborating and excited about the solutions.

How did ARC Create the Images?

ARC staff used GPT-4, an advanced language model available via ChatGPT, to create the images. We developed a detailed prompt for each set of images, and then worked within ChatGPT to revise them as needed. We found that treating ChatGPT like a coworker and providing feedback in a conversational style yielded the best results. We also had ARC staff that live in Rajkot closely review the draft images to ensure they were culturally relevant and accurately represented the focus issue.

As with any new technology, we experienced some obstacles. The AI frequently included details in the images that were rooted in cultural biases. For example, initial versions of images 1 and 2 showed women exclusively dressed in traditional attire, while men wore modern clothing. In addition, both men and women looked much older than the typical working population in Rajkot. ARC Rajkot staff reviewed the images closely to identify issues, and we used specific prompts to instruct ChatGPT to revise the images to reflect people in the community more accurately.

What’s Next?

Our experience in Rajkot taught us important lessons on the opportunities and limitations of AI. While it took time to create images that were inclusive of all groups living in Rajkot and accurately portrayed their concerns, the final images allowed ARC staff and residents to have more meaningful discussions about perspectives and goals. With human understanding of the drawbacks, AI can enhance ARC and other urban resilience initiatives’ efforts.

Asia Resilient Cities (ARC) is a five-year cooperative agreement funded by the U.S. Agency for International Development (USAID) under Agreement No. 7200AA22CA00011, beginning September 23, 2022. ARC is implemented by JSI with partners BRAC, DevTech Systems, Inc., ICF, International Organization for Migration, John Snow India Private Limited, and The Asia Foundation. This product is made possible by the generous support of the American people through USAID. Its contents are the responsibility of ARC and do not necessarily reflect the views of USAID or the U.S. Government.