[1L3-GS-9a] Human interface

2:55 PM - 3:10 PM JST(5:55 AM - 6:10 AM UTC)

[1L3-GS-9a-06] An Idea Suggestion Method Using Generative AI to Support Ideation in SF Prototyping

Misuzu Ando¹, Atsuya Fujimoto¹, Mengsi Liu¹, 〇Hirotaka Osawa¹ (1. Faculty of Science and Technology, Keio University)

Keywords:

SF Prototyping,Science Fiction

This study targeted a short SF writing task in SF prototyping (SFP) to clarify the effects of manipulating the timing of AI-generated story idea presentation on AI anchoring, group diversity, work evaluation, and self-reported AI usage. A face-to-face within-subjects experiment (N=24, three themes) was conducted with three conditions: no AI support (C0), AI idea presented before writing (C1), and AI idea presented after independent thinking (C2). Condition order was counterbalanced via Williams design. Semantic similarity and within-condition convergence were measured using text-embedding-3-large. C1 exhibited significantly higher group diversity than C0 (STM difference = 5.89, p = 0.021), contrary to the diversity loss anticipated by prior work. No significant differences were found in anchoring or work evaluation after correction. The AI usage questionnaire showed a trend of lower 'idea starting point' ratings in C2, though not significant after Holm correction. These results suggest that in short SF writing, AI story ideas may serve as divergent stimuli rather than convergent anchors, and that diversity metrics are sensitive to task themes and participant attributes.