Human-Centered AI for Climate Adaptation
This project aims to develop a user-friendly framework using Visual Analytics (VA) and Explainable Artificial Intelligence (XAI) to create effective climate adaptation plans across different sectors. Many water resources managers face so-called 'Deep Uncertainty', where key information for decision-making is uncertain or disputed. This framework will help in designing robust adaptation strategies that avoid irreversible commitments and identify critical decision points.
Using advanced AI techniques like multi-objective evolutionary algorithms and reinforcement learning, the project will generate adaptive pathways that surpass human-developed strategies in test scenarios. The challenge lies in making these AI-driven policies transparent and understandable for decision-makers, akin to challenges in healthcare and finance.
The project will pioneer interactive VA tools tailored for climate adaptation, building on recent developments in XAI. Starting with a test case in the Netherlands, the system will simulate flood planning along the river Waal, balancing urban, agricultural, and shipping interests. Decision-makers can interact with the system as it explores various adaptation options, providing insights into how climate data informs decision triggers.
Led by Dr. David Gold and Dr. Angelos Chatzimparmpas, the project aims to create an open-source VA system for dynamic climate adaptation pathways. Results will be disseminated through open-access publications and showcased at key industry conferences, fostering broader adoption of AI in climate resilience planning.