DESIGNING TRUST: HOW SYSTEM-LEVEL DESIGN ATTRIBUTES SHAPE AI-ASSISTED DESIGN IDEATION

Recent advancements in artificial intelligence (AI) have enabled the use of conversational agents in engineering design, particularly in early-stage concept generation. While AI’s technical capabilities continue to improve, less is known about how system-level design attributes influence trust in AI systems and user interaction. Trust has emerged as a critical factor in human-AI collaboration, shaping whether and how users engage with AI systems. This study proposed and developed the AI Design Assistant (AIDA), a chatbot-based AI system that supports creative ideation with contextualized, criteria-aligned interaction. Using AIDA, the study investigates how three target design attributes—appearance, system performance, and empathetic behavior—affect user trust and post-interaction perceptions. Through a controlled experimental framework, the preliminary data analysis shows that performance and empathy significantly enhance trust, while individual differences, such as users’ initial trust and age, further modulate trust development. Notably, empathetic behavior improved perceived effort, enjoyment, and usefulness, even when system performance was low, suggesting that emotional connection can partially offset technical limitations. Conjoint analysis confirmed the impact of high-level design features but indicated lower relative importance for empathy, highlighting the complementary value of combining statistical and preference-based approaches. These findings offer early insights into the design of emotionally intelligent, trustworthy AI systems for creative domains.