Integrating Recognition-Primed Decision Making with LLMs for Expert-Like Wildfire Response

Kamsi E. Amaeshi
Prof. Nazmul Shahadat, Faculty Mentor

The increasing frequency and severity of natural disasters have intensified the need for effective decision-making under conditions of uncertainty, time pressure, and rapidly evolving information. While recent advances in artificial intelligence (AI) have improved data processing, coordination, and planning in disaster response, existing systems remain limited. This research explores integrating cognitive science, specifically the Recognition-Primed Decision (RPD) model, with large language models (LLMs) to create an intelligent agent capable of adaptive wildfire response.

Grounded in the principles of Naturalistic Decision Making (NDM), the agent simulates expert reasoning through structured prompts that engage pattern recognition, anomaly detection, and mental simulation. The agent operates within a custom wildfire simulation environment built using Mesa, and leverages historical case data as memory to inform context-sensitive decisions.

This work contributes to cognitively grounded AI for emergency response, offering a hybrid framework that bridges human decision theory, language-based reasoning, and interactive simulation

Keywords: Recognition-Primed Decision Model, Naturalistic Decision Making, LLMs, Crisis Decision-Making, Situational Awareness, Wildfire Response , Expert Behavior Modeling

Topic(s):Computer Science
Psychology

Presentation Type: Oral Presentation

Session: -
Location: MG 2007
Time: 

Add to Custom Schedule