Can a troubleshooting AI assistant improve task performance in industrial contexts?
通过实地实验，研究了AI故障排除助手在火车调试这一知识密集型任务中的作用，发现AI辅助显著提升了任务绩效，尤其对经验较少的技术人员帮助更大，且效果受技术人员对AI的态度和熟悉度影响。
Access to domain expertise is critical to problem-solving activities. This study investigates the role and usefulness of an AI-based troubleshooting assistant in the knowledge-intensive context of train commissioning. The authors developed a multilingual, user-centred chatbot using a Retrieval-Augmented Generation framework and a Large Language Model to provide real-time, project-specific troubleshooting support. A controlled field experiment with 19 commissioning technicians completing 173 tasks was conducted to evaluate the effect of the AI assistant. Results show that AI-assisted users significantly outperformed non-users in task performance. The benefits were more substantial among less experienced technicians, which emphasises the prospect of AI assistants in bridging skill gaps. Performance gains were also moderated by the AI attitudes and AI familiarity of technicians, suggesting that organisations should take these factors into account when striving to adopt AI assistants. The findings provide empirical evidence for the role of generative AI in scaling operational knowledge, enhancing worker performance, and improving the efficiency of engineering workflows. This study contributes to the discussion of the role of generative AI in problem-solving by demonstrating a novel application for frontline decision support in complex, high-variance industrial tasks.