AI for Creativity: A GenAI-Based Approach for Early Concept Design and Its Impact on Senior Architects

Paper Title

AI for Creativity: A GenAI-Based Approach for Early Concept Design and Its Impact on Senior Architects

Publication Info

• Topic area: Application of Generative AI (GenAI) in early-stage architectural concept design for senior professionals.
• Keywords: Generative AI, architectural design, early concept design, senior architects, human–AI collaboration, creativity support, professional fulfillment, iterative design, system usability, dynamic role flexibility.

Background and Problem

• Problem / challenge: Existing GenAI tools primarily focus on execution phases and offer limited support for the early concept design stage, which is critical for senior architects. These tools often lack iterative exploration and fail to align with senior professionals' workflows and cognitive needs, potentially undermining creativity and professional autonomy.
• Significance: Addressing these gaps is crucial for enhancing architectural innovation, supporting senior architects' leadership roles, and advancing human–AI collaboration in creative fields.
• Motivation and related work: Prior research has explored GenAI's potential in visualization and iterative refinement but has largely overlooked the needs of senior professionals in early-stage design. Existing tools often prioritize efficiency over deeper conceptual exploration, leaving a gap in supporting senior architects' iterative and reflective design processes.

Solution

• Proposed approach: Development of EarlyArchi, a GenAI-driven system tailored to senior architects for early-stage concept design, integrating structured workflows, iterative refinement, and customizable evaluation.
• Novelty:
  1. A user-centered design pipeline specifically addressing senior architects' workflows and cognitive needs.
  2. Integration of GenAI for both concept generation and evaluation within a single platform.
  3. Introduction of a Dynamic Role Flexibility framework, outlining adaptive human–GenAI collaboration modes.
  4. Empirical insights into how GenAI impacts senior architects' expertise, competency, and fulfillment.
• Procedure and key techniques:
  • Conducted a formative study (N=11) to identify senior architects' challenges and derive design goals.
  • Developed EarlyArchi with three core modules (input, AI processing, output) and aligned it with the RIBA Plan of Work and Double Diamond Design Model.
  • Evaluated EarlyArchi through a user study (N=13), combining quantitative surveys and qualitative interviews to assess creativity support, usability, and professional impact.

Results

• Concrete findings:
  • Participants rated EarlyArchi positively for organizing ideas (average 4.54/5) and facilitating communication (average 4.69/5).
  • Mixed satisfaction with GenAI outputs: creativity support ratings averaged 3.46/5 for "results worth effort" and 3.46/5 for output quality.
  • Enhanced expertise (average 5.31/6 for knowledge improvement) and competency (average 4.85/5 for job performance), but limited impact on leadership skills (average 3.46/5).
  • Moderate fulfillment ratings, with participants appreciating efficiency gains but expressing concerns about GenAI's controllability and creative limitations.
• Advantage over baselines: EarlyArchi integrates concept generation and evaluation in a single platform, reducing fragmentation and enabling iterative refinement, which existing tools lack.
• Experiments / evaluation:
  • User study with 13 senior architects, assessing creativity support (Creative Support Index), usability (Technology Acceptance Model 3), and professional impact (Kirkpatrick Model, Professional Fulfillment Index).
  • Mixed-method analysis combining surveys and thematic interviews.
• Limitations and future work:
  • Current reliance on general GenAI models limits domain-specific precision.
  • Small sample size and remote study setup may affect generalizability.
  • Future work includes developing domain-specialized AI models, refining prompt guidance, and conducting longitudinal studies in diverse architectural contexts.

Summary

This paper introduces EarlyArchi, a GenAI-based system designed to support senior architects in early concept design by integrating structured workflows, iterative refinement, and customizable evaluation. A user study (N=13) demonstrated that EarlyArchi enhances expertise, competency, and efficiency, though participants expressed mixed satisfaction with GenAI outputs and concerns about controllability. The study also identified three modes of human–GenAI collaboration—fully AI-driven, GenAI-led, and human-led—emphasizing the need for dynamic role flexibility. These findings offer actionable insights for designing future GenAI tools that balance automation with expert-driven creativity, addressing the nuanced needs of senior professionals in architecture and other creative domains.