AI-driven decision-making for water resource planning and hazard mitigation using automated multi-agents

Abstract

This project simulates the Multi-Hazard Tournament (MHT) framework, a decision support system designed for the U.S. Army Corps of Engineers, using AI agents to enhance decision-making processes for flood mitigation and water resource management. Traditional hydrological models often overlook social dynamics and community preferences crucial for sustainable implementation. The objective of the framework is to develop optimal strategies for protecting water resources, habitats, and communities within a defined budget. The simulation integrates AutoGen for managing multi-agent interactions and DarkIdol-Llama-3.1-8B, an advanced language model, to facilitate complex, long-context discussions. AI agents are configured with distinct roles and engage in structured dialogues to collaboratively evaluate and refine mitigation strategies. Analysis of 1,000 diverse agents revealed age as the most significant factor (importance: 0.14) influencing budget allocation, with younger participants (19-30) favoring immediate infrastructure investments while older participants (61+) preferred conservative strategies. The study demonstrates the potential of AI-driven simulations to replicate real-world collaborative environments, improving stakeholder engagement and enhancing the efficiency of hazard mitigation planning. The findings highlight the effectiveness of AI agents in multi-stakeholder decision-making processes, offering valuable insights for disaster risk reduction. This work contributes significantly to fostering more resilient, well-prepared communities through innovative approaches to decision-making.