一句话：两阶段调查揭示软件工程研究者对在研究工作流中使用 LLM 的态度、接受边界与科研诚信关切。
本文关注软件工程（SE）研究者如何将 AI 工具尤其是大语言模型（LLM）融入研究工作流，以及由此涉及的负责任使用与科研诚信问题。作者在一所大型独立欧洲研究机构开展两阶段问卷调查，考察研究者对 LLM 在新兴实证 SE 实践中的规范与态度。结果显示，研究者更倾向让 LLM 参与范围狭窄、可验证的小任务，并将其作为传统方法的补充工具；相反，用它评估他人工作被认为不合适。对于传统上体现真实人力投入与情感承诺的高风险任务，研究者态度存在分歧。
This two-phase survey study investigates Software Engineering (SE) researchers' attitudes toward incorporating AI tools, particularly large language models (LLMs), in their workflows. Conducted at a large, independent European research institute, the study provides insights into the emerging norms and attitudes regarding AI integration within empirical SE practices, crucial insights for ensuring responsible AI use and maintaining scientific integrity in the SE field. Our findings reveal that SE researchers prefer LLM contributions when applied on small, narrowly scoped, and verifiable tasks rather than open-ended tasks, and used as a as supplementary tool to traditional methods. In contrast, it is viewed as inappropriate for evaluating others' work. Finally, we observed conflicting views regarding high-stakes tasks traditionally reflecting genuine human effort and emotional commitment.
@inproceedings{paperbot8939,
title = {Attitudes Towards LLM Use Among Software Engineering Researchers: Results From A Two-Phase Survey Study},
author = {Viggo Tellefsen Wivestad and Astri Moksnes Barbala},
booktitle = {FSE},
year = {2025},
doi = {10.1145/3696630.3731671}
}