Date of Award

5-12-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Information Science

First Advisor

Serhan Dagtas

Abstract

Large Language Models (LLMs) have demonstrated significant potential in disaster-response decision support, however, their deployment in high-stakes humanitarian settings raises critical concerns regarding factual reliability, fairness, temporal validity, and governance compliance. Hallucinated outputs, demographic bias, and outdated recommendations can directly impact vulnerable populations and undermine public trust. This dissertation proposes a governance-aware multi-agent framework designed to enhance fairness and temporal accuracy in disaster-response systems through structured Retrieval-Augmented Generation (RAG), verification-driven orchestration, and adaptive correction mechanisms.The proposed architecture decomposes response generation into specialized agents responsible for real-time retrieval, fact-checking, bias auditing, temporal validation, threshold-based correction, and monitoring. By embedding governance constraints directly into the generative lifecycle, the system operationalizes measurable dimensions of factuality, fairness, temporal freshness, and latency. A reinforcement learning–optimized orchestration layer further enables adaptive triggering of validation modules based on uncertainty signals, balancing reliability and efficiency. Evaluation is conducted using controlled disaster-response scenarios and domain-adapted fine-tuning on crisis benchmark datasets. Results demonstrate substantial improvements over a standalone baseline model, including significant gains in factual grounding, equitable representation across demographic contexts, and recency alignment of responses. While multi-agent orchestration introduces latency overhead, adaptive validation strategies mitigate performance trade-offs while preserving governance guarantees.This work contributes a scalable and principled framework for trustworthy AI deployment in disaster-response environments, integrating technical robustness with ethical and regulatory alignment. By coupling fairness auditing with temporal reasoning in a coordinated multi-agent architecture, the proposed system advances the development of accountable, resilient, and humanitarian-centered large language model applications.

Download

Share

COinS