Prototype of a Generative AI–Based Analogy Application for Human Error Case Analysis

Abstract

Research at the intersection of human factors analysis and large language models (LLMs) has grown rapidly in recent years; however, much of this work emphasizes automation and efficiency, evaluating success primarily through model-centric metrics. In contrast, this study reframes generative AI not as an automation tool for analysis but as a collaborative partner for cognitive stimulation, and proposes PromptWeave, a prompt-design methodology intended to expand, deepen, and transform an analyst’s reasoning. We applied PromptWeave to industrial accident cases and conducted a quantitative evaluation using human-centered KPIs. The results indicate consistently high performance across all KPIs, supporting the utility of PromptWeave as a reproducible collaboration protocol executable on an LLM platform.