Agentic LLM Framework for Adaptive Decision Discourse

TL;DR

This paper presents an agentic LLM framework that simulates stakeholder dialogues to generate adaptive, equitable decision recommendations in complex, uncertain scenarios, demonstrated through flood response case studies.

Contribution

Introduces a novel agentic LLM framework for simulating decision discourse with diverse stakeholder personas, enabling scalable, context-aware recommendations for complex challenges.

Findings

Framework produces balanced recommendations considering social, economic, and environmental factors.

Simulated flood response scenarios demonstrate adaptability and robustness of the approach.

Potential for scalable decision support in high-stakes, uncertain environments.

Abstract

Effective decision-making in complex systems requires synthesizing diverse perspectives to address multifaceted challenges under uncertainty. This study introduces an agentic Large Language Models (LLMs) framework for simulating decision discourse - the deliberative process through which actionable strategies are collaboratively developed. Unlike traditional decision-support tools, this framework simulates diverse stakeholder personas, each bringing unique priorities, expertise and value-driven reasoning to a dialogue that emphasizes trade-off exploration in a self-governed assembly. We present explorative results fostering robust and equitable recommendations, with two use cases: first, our framework simulates a response to the floods that occurred on July 2025 in Texas; second, a hypothetical extreme flooding in a Midwestern township under varying forecasting uncertainty. Recommendations made balance competing priorities considered through social, economic and environmental dimensions, setting a foundation for scalable and context-aware recommendations and transforming how decisions for real-world high-stake scenarios can be approached in digital environments. This research explores novel and alternate routes leveraging agentic LLMs for adaptive, collaborative, and equitable recommendations, with implications across domains where uncertainty and complexity converge.