Water On My Block: Reflections on Building A Participatory Artificial Intelligence System For Precision Weather With Scientists and An Urban Community

Paper Title

Water On My Block: Reflections on Building A Participatory Artificial Intelligence System For Precision Weather With Scientists and An Urban Community

Publication Info

• Topic area: Participatory AI for hyper-local climate modeling in urban communities.
• Keywords: Participatory AI, hyper-local climate models, precision weather, community engagement, urban flooding, green infrastructure, HCI, data governance, citizen science, climate services.

Background and Problem

• Problem / challenge: Existing climate AI systems lack the granularity to model weather impacts at the neighborhood level, limiting their utility for urban communities. There is minimal practical guidance on how participatory AI can be implemented to bridge the gap between scientists and communities.
• Significance: Hyper-local climate data can help underserved urban communities mitigate extreme weather impacts, advocate for resources, and make informed decisions about green infrastructure.
• Motivation and related work: Prior work has explored participatory design and citizen science but has not extensively studied real-world participatory AI systems in urban settings. Existing efforts, such as deploying air quality sensor networks, provide a foundation but leave gaps in addressing community-specific needs and sustained engagement.

Solution

• Proposed approach: The authors developed a participatory AI framework to co-design and implement a hyper-local climate AI system, focusing on flood reporting and community advocacy.
• Novelty:
  1. Introduced the term "precision weather" for hyper-local climate AI models using neighborhood-level data.
  2. Developed the Water On My Block app to collect flood reports and support both scientific modeling and community advocacy.
  3. Employed a participatory design process involving scientists and community members through interviews, workshops, and co-design sessions.
  4. Addressed challenges in data governance, privacy, and trust-building between stakeholders.
• Procedure and key techniques:
  1. Conducted four research phases: relationship building, interviews, Community Cafes, and app development.
  2. Gathered input from 15 scientists and 48 community members through interviews and workshops.
  3. Co-designed a flood reporting app with features like a map interface, community feed, and advocacy tools.
  4. Transferred app ownership to the community partner after iterative evaluation and refinement.

Results

• Concrete findings:
  • 13/14 participants in the final evaluation found the app's community feed valuable.
  • 55% of residents preferred a flood reporting tool over other app prototypes.
  • Scientists valued the app's potential for collecting ground truth data to improve AI models.
• Advantage over baselines: The app bridged the gap between scientific data needs and community priorities, creating actionable outcomes for both groups.
• Experiments / evaluation:
  • Conducted three Community Cafes to gather feedback on app design and functionality.
  • Evaluated the app with scientists and community members, addressing usability, privacy, and data governance concerns.
• Limitations and future work:
  • The app's integration with scientific AI models was not fully evaluated due to the timeline.
  • Future work could explore long-term sustainability, deeper community involvement in AI pipelines, and scaling participatory AI methods.

Summary

This paper presents a three-year case study on participatory AI for hyper-local climate modeling, focusing on urban flooding in Chicago's Chatham neighborhood. The authors co-designed the Water On My Block app, enabling residents to report flooding and advocate for resources while providing scientists with valuable data for precision weather AI models. The study highlights the importance of trust-building, data governance, and actionable outcomes in participatory AI. While the app successfully aligned stakeholder needs, challenges remain in scaling such approaches and ensuring long-term sustainability. This work contributes practical insights for HCI researchers, scientists, and policymakers aiming to develop equitable, community-centered AI systems.