Paper Type
Complete
Paper Number
PACIS2026-1745
Description
Design thinking (DT) increasingly incorporates artificial intelligence (AI), yet most research treats AI as a support tool rather than a collaborative teammate. This study investigates how replacing human roles with role-based AI agents influences collaboration in DT workshops. Drawing on role theory and framing theory, we compare AI and human actors in two key roles: facilitator and participant. A within-subject experiment with 30 participants across 10 counterbalanced workshops compared a human-only condition with an AI-assisted condition using GPT-4–based chatbots on Discord. Participants evaluated collaboration efficiency, cognitive load, and trust after each session. Results show that both AI facilitators and AI participants significantly improved collaboration efficiency compared to humans. However, human facilitators and participants were trusted more, and no significant differences were found in cognitive load. These findings suggest that AI enhances procedural efficiency in DT collaboration but does not automatically achieve the social legitimacy and trust associated with human actors
Recommended Citation
Indra Widi Tamtama, Gabriel; I-Hsuan, Chen; Wang, Jyun-Cheng; Figueroa, Roberto B.; and Budi Santoso, Halim, "From Tool to Teammate: Understanding Role-based AI-enabled Design Thinking" (2026). PACIS 2026 Proceedings. 9.
https://aisel.aisnet.org/pacis2026/ai_fow/ai_fow/9
From Tool to Teammate: Understanding Role-based AI-enabled Design Thinking
Design thinking (DT) increasingly incorporates artificial intelligence (AI), yet most research treats AI as a support tool rather than a collaborative teammate. This study investigates how replacing human roles with role-based AI agents influences collaboration in DT workshops. Drawing on role theory and framing theory, we compare AI and human actors in two key roles: facilitator and participant. A within-subject experiment with 30 participants across 10 counterbalanced workshops compared a human-only condition with an AI-assisted condition using GPT-4–based chatbots on Discord. Participants evaluated collaboration efficiency, cognitive load, and trust after each session. Results show that both AI facilitators and AI participants significantly improved collaboration efficiency compared to humans. However, human facilitators and participants were trusted more, and no significant differences were found in cognitive load. These findings suggest that AI enhances procedural efficiency in DT collaboration but does not automatically achieve the social legitimacy and trust associated with human actors
Comments
02-FutureofWork