Large language models (LLMs) offer disaster risk management (DRM) practitioners a promising “plug-and-play” solution for information extraction on online media data. However, they can struggle with the strict precision required for identifying semantically and structurally complex disaster causes. We evaluated how off-the shelf LLMs extract causal information from flood-related texts and found a strong recall bias. To diagnose these errors, our primary contribution is a comprehensive, human-annotated qualitative error taxonomy using inductive and deductive coding. We identified specific linguistic traps, i.e., impact descriptions, background context, and subjective opinions, trigger over-extraction, indicating models have a patterned failure. As an exploratory second step, we apply this taxonomy to design an ablation study of prompt-based interventions using definitions and exclusions. Our results show that our approach successfully shifts models along the precision–recall trade-off, i.e., curbing over-extraction, improving precision and domain-specific F0.5-score for recall-biased models. We demonstrate that by leveraging qualitative error analysis, domain experts can not only characterize and refine off-the-shelf LLMs, but also navigate the tradeoffs of prompting.

The adoption of large language models (LLMs) in high-stake scenarios continues to be a challenge due to lack of effective confidence calibration. Although LLMs are capable of providing convincing self-explanations and verbalizing confidence in NLP tasks, they tend to exhibit overconfidence when using generative or free-text rationales (e.g. Chain-of-Thought), where reasoning steps tend to lack verifiable grounding.In this paper, we investigate whether adding explanations in the form of extractive rationales –snippets of the input text that directly support the predictions, can improve the confidence calibration of LLMs in classification tasks.We examine two approaches for integrating these rationales: (1) a one-stage rationale-generation with prediction and (2) a two-stage rationale-guided confidence calibration.We evaluate these approaches on a disaster tweet classification task using four different off-the-shelf LLMs. Our results show that extracting rationales both before and after prediction can improve the confidence estimates of the LLMs. Furthermore, we find that replacing valid extractive rationales with irrelevant ones significantly lowers model confidence, highlighting the importance of rationale quality.This simple yet effective method improves LLM verbalized confidence and reduces overconfidence in possible hallucination.

Large Language Models (LLMs) have shown promise in automating high-labor data tasks, but the adoption of LLMs in high-stake scenarios faces two key challenges: their tendency to answer despite uncertainty and their difficulty handling long input contexts robustly.We investigate commonly used off-the-shelf LLMs’ ability to identify low-confidence outputs for human review through “check set selection”–a process where LLMs prioritize information needing human judgment.Using a case study on social media monitoring for disaster risk management,we define the “check set” as a list of tweets escalated to the disaster manager when the LLM has the least confidence, enabling human oversight within budgeted effort.We test two strategies for LLM check set selection: individual confidence elicitation – LLMs assesses confidence for each tweet classification individually, requiring more prompts with shorter contexts, and direct set confidence elicitation – LLM evaluates confidence for a list of tweet classifications at once, using less prompts but longer contexts.Our results reveal that set selection via individual probabilities is more reliable but that direct set confidence merits further investigation.Direct set selection challenges include inconsistent outputs, incorrect check set size, and low inter-annotator agreement. Despite these challenges, our approach improves collaborative disaster tweet classification by outperforming random-sample check set selection, demonstrating the potential of human-LLM collaboration.