一句话：提出 EAI-USG 工具，利用 LLM 与 RAG 自动将高层伦理 AI 需求转化为敏捷工作流中的伦理用户故事。
尽管 AI 伦理指南的讨论广泛，但软件开发实践中常缺乏可操作落地。本文提出 EAI-USG，一个利用大语言模型与检索增强生成（RAG）的工具，将高层的伦理 AI 需求自动翻译为“伦理用户故事”，从而把伦理原则转化为敏捷工作流中的具体任务，并生成结构化故事。初步评估邀请 30 名软件从业者参与，结果显示他们普遍认为生成的故事清晰且有用，该方案能在需求工程阶段以较小的额外工作量促进伦理的可操作化。
While the subject of ethical guidelines for AI is widely discussed, there is often a lack of practical application in the context of software development. In order to address this issue, we present EAI-USG, a tool that automatically translates high-level ethical requirements into Ethical User Stories, transforming ethical principles into concrete tasks within agile workflows. Powered by large language models with retrieval-augmented generation, the tool produces structured stories that practitioners perceive as clear and useful. A preliminary evaluation involving 30 software professionals suggests that this approach facilitates the operationalization of ethics during requirements engineering with only a minor additional effort.
@inproceedings{paperbot11581,
title = {Bridging Theory and Practice: Ethical AI User Story Generator - Translating Ethical AI Requirements into Ethical User Stories},
author = {João Rossi and Edna Canedo and Pekka Abrahamsson and Jose Siqueira de Cerqueira and Fabiana Freitas Mendes},
booktitle = {ICSE},
year = {2026},
doi = {10.1145/3774748.3795665}
}