文献类型：会议论文

英文题名：Do Designers Generate Ideas Better with a Team of Specialized AI Agents or a General AI Agent?

作者：Chen, Yuxiang[1];Chen, Yue[1];Han, Yu[1];Zhou, Yudi[1]

机构：[1]East China Univ Sci & Technol, Shanghai 200237, Peoples R China

会议论文集：2025 International Conference on Cross Cultural Design-CCD-Annual

会议日期：JUN 22-27, 2025

会议地点：Goeteborg, SWEDEN

语种：英文

外文关键词：The team of specialized AI agents; A general AI agent; Team Creativity; idea generation; collaborative design

摘要：Many designers are adopting generative AI for idea generation and brainstorming, such as ChatGPT. Such AI possesses broad adaptability and the capability to execute diverse tasks but lacks specificity tailored to the demands of different tasks in the design process, such as brainstorming, knowledge searching, and idea evaluation. To support these different tasks, this study applied team role theories and proposed three specialized AI agents, namely creator, expert, and doer, with distinct roles and personalities. To confirm the effectiveness of specialized AI agents on ideation, we conducted a within-group experiment of 15 undergraduate students and compared the team of specialized AI agents with a general AI agent. We found that the team of specialized AI agents induced more positive and dominant emotions and increased trust, satisfaction, and overall team performance. This study integrates team role theories into AI agent design, demonstrating how a team of AI agents with specialized roles can enhance designers' idea generation. It also offers a role-specialized approach for designing AI agents to support designers' idea generation.