ScholarWorks@성균관대학교

초록

While AI text generation is instantaneous, image generation and reasoning models take longer. To signal processing time to users, platforms tend to use interface cues such as progress bars, skeleton screeners, and textual descriptions of the generation process. However, little is known about whether and how these cues affect user experience of generative AI tools. Through semi-structured interviews with 11 participants who interacted with various AI image generation tools, we examined user perceptions of waiting times and progress indicators, which we term “generation cues.” Data show that users are generally accepting of, sometimes even value, waiting times, viewing them as an inherent part of the creative process. However, they rarely notice generation cues. These findings challenge traditional usability principles of speed, efficiency, and feedback, suggesting the need to rethink waiting times and generation cues in the context of generative AI. We offer design recommendations for leveraging waiting times and generation cues to enhance human-AI collaboration in creative domains. © 2025 Copyright held by the owner/author(s).

키워드