Creativity from Surprise: Bridging the Gap Between Fashion Designers' Inspiration Work and AI Creative Support Tools

Paper Title

Creativity from Surprise: Bridging the Gap Between Fashion Designers' Inspiration Work and AI Creative Support Tools

Publication Info

• Topic area: Generative AI as a creativity support tool in fashion design.
• Keywords: Generative AI, fashion design, creativity support tools, surprise, co-creation, early-stage ideation, human-centered AI, visualization tools, design innovation, strategic surprise.

Background and Problem

• Problem / challenge: Current Generative AI tools often fail to produce meaningful and usable surprises in early-stage fashion ideation due to limitations in interpretability, control, and contextual sensitivity.
• Significance: Harnessing AI-generated surprise could help designers overcome creative fixation, explore novel directions, and maintain brand identity while innovating within commercial constraints.
• Motivation and related work: Prior studies have explored AI interfaces for aligning designer intent with outputs, but lack empirical insights into how fashion designers define and utilize surprise strategically. This paper addresses the gap by studying the role of surprise in fashion ideation and proposing actionable system features for GenAI tools.

Solution

• Proposed approach: A methodological blueprint for GenAI-powered visualization tools that deliver strategic, controllable, and context-aware surprise to support co-creative workflows in fashion design.
• Novelty:
  1. Reframing surprise as a designable mechanism for co-creative interaction in GenAI tools.
  2. Empirical insights into how fashion designers use AI-generated surprise in early-stage ideation.
  3. Actionable system features to enable user-aligned, controllable, and context-aware surprise.
  4. A four-stage framework for iterative co-creative engagement: setting expectations, supporting interaction, recovering from errors, and improving over time.
• Procedure and key techniques:
  • Phase 1: Semi-structured interviews with 20 professional fashion designers to understand surprise in ideation.
  • Phase 2: Design workshop with 12 graduate students to explore system features for GenAI tools.
  • Analysis of interview and workshop data to identify themes and propose interaction features.

Results

• Concrete findings:
  • GenAI-driven surprise manifests as Internal Surprise (cognitive disruption) and External Surprise (market or cultural insights).
  • Current GenAI tools excel in visualization speed, design integration, communication, and bold visual effects but struggle with specificity, understanding abstract concepts, associative expansion, and iterative consistency.
  • Proposed system features include multi-language support, structured prompting, real-time feedback, error recovery mechanisms, and long-term workflow tracking.
• Advantage over baselines:
  • Strategic surprise aligns AI outputs with user goals, enabling meaningful novelty rather than random or incoherent results.
  • Enhanced control and contextual grounding improve usability and creative agency compared to current GenAI tools.
• Experiments / evaluation:
  • Interviews with 20 designers (mean experience: 11.15 years) and a workshop with 12 design students (mean age: 29.42 years) provided qualitative insights into GenAI capabilities and limitations.
  • Wireframes and thematic analysis informed the proposed design blueprint.
• Limitations and future work:
  • Findings are limited to the fashion domain and predominantly Asian sample.
  • Workshop participants were students, not industry professionals, which may emphasize exploratory creativity over practical constraints.
  • The blueprint has not been implemented or validated in a real-world prototype.
  • Future research should test applicability in other design domains and evaluate prototypes quantitatively.

Summary

This study investigates how fashion designers conceptualize and utilize surprise in early-stage ideation and proposes a methodological blueprint for GenAI-powered visualization tools to deliver strategic, controllable, and context-aware surprise. Through interviews and a design workshop, the paper identifies key designer needs and limitations of current GenAI tools, such as lack of specificity and iterative consistency. Proposed features include structured prompting, real-time feedback, and long-term workflow tracking. While focused on fashion design, the findings and blueprint have potential applicability across creative domains. Future work should implement and evaluate these features in professional settings to assess their impact on creative diversity and efficiency.