Generative AI Advances Architectural Design for Novice Designers, Showing Performance Gains with 36 Participants

The increasing prevalence of generative artificial intelligence (GenAI) prompts crucial questions about its impact on creative work, particularly in fields like architecture. Han Jiang, Yao Xiao, and Rachel Hurley, from Worcester Polytechnic Institute, alongside Shichao Liu, investigated how GenAI affects performance, creative self-belief and mental effort during the initial stages of architectural design. Their research, involving a comparative study of student designers, reveals a nuanced relationship between AI assistance and design outcomes. While GenAI did not universally enhance performance, the study demonstrates significant improvements for those with limited prior experience, suggesting a potential role in democratising design processes. However, the findings also indicate a decline in designers’ confidence in their own creative abilities when utilising these tools, highlighting a need to understand the psychological effects of integrating AI into creative workflows.

Architectural conceptual design tasks were investigated with thirty-six student participants drawn from Architectural Engineering and other disciplines. Participants completed a two-phase architectural design task, initially working independently before undertaking the same task with external tools. The study employed a GenAI-assisted condition and a control condition, utilising an online repository of existing architectural projects. Design outcomes were evaluated by expert raters, with self-efficacy and cognitive load measured via self-report questionnaires after each phase. Difference-in-differences analyses revealed no overall performance advantage for GenAI across all participants; however, subgroup analyses demonstrated that GenAI significantly improved design performance specifically for novice designers. In contrast, general creativity….

AI’s Impact on Design Thinking Skills

The study examines the influence of generative artificial intelligence (AI) on design thinking processes, with particular emphasis on self-efficacy, learning experiences, and the development of higher-order thinking skills. It finds that students who used AI tools reported enhanced confidence in their design abilities, especially when AI-generated outputs allowed them to visualize and validate their ideas. This tangible feedback strengthened their belief in their creative potential and contributed to a more engaging and interactive learning experience. Real-time feedback from AI tools supported iterative refinement, making the learning process more dynamic and responsive.

The findings further indicate that AI use promotes higher-order thinking skills, including critical thinking, problem-solving, and innovation. By encouraging students to move beyond conventional design approaches, AI tools fostered creative and flexible problem-solving strategies. However, the study also highlights a nuanced impact on design fixation and divergent thinking. While AI can help reduce fixation by offering alternative perspectives, excessive reliance on AI may limit idea diversity. Students who were encouraged to generate multiple ideas independently before engaging with AI demonstrated stronger divergent thinking abilities, underscoring the importance of balanced AI integration.

Self-regulation emerged as a key factor in effectively leveraging AI for design tasks. Students with clear goals and the ability to manage their AI usage showed higher motivation and engagement. Methodologically, the study employed surveys to measure self-efficacy and perceived learning experiences, interviews to gather qualitative insights, and observations during AI-integrated design workshops. Quantitative data were analyzed using statistical techniques such as ANOVA and regression analysis, while qualitative data were coded to identify recurring themes related to AI-supported design learning.

The implications of the study suggest that educators should integrate AI tools in ways that complement rather than replace traditional design methods. Curriculum design should incorporate prompt engineering, critical thinking, and hybrid human–AI learning approaches to support self-regulated learning. While generative AI offers significant benefits for enhancing creativity and learning outcomes, the study emphasizes the need for careful and balanced use to avoid over-reliance. Future research is encouraged to explore the long-term effects of AI on design education and to examine how different types of AI tools influence various dimensions of design thinking.

AI Boosts Novice Designers, Lowers Self-Efficacy

The research team investigated the influence of generative AI on architectural conceptual design, meticulously examining performance, creative self-efficacy, and cognitive load in thirty-six student participants. Participants completed a two-phase design task, initially working independently and then utilising either generative AI tools or an online repository as a control. Subsequent difference-in-differences analyses revealed no overall performance benefit from using generative AI across the entire participant group, however, detailed subgroup analysis demonstrated a significant improvement in design performance specifically for novice designers. This suggests the technology’s utility is contingent on existing skill levels.

Experiments recorded a decline in general creative self-efficacy amongst students employing generative AI during the design process, a noteworthy finding regarding the psychological impact of these tools. Despite this, measurements of cognitive load did not show significant differences between the conditions, indicating that the introduction of AI did not inherently increase mental strain. However, analysis of usage patterns revealed a correlation between iterative idea generation and visual feedback, and a corresponding reduction in cognitive load, highlighting the importance of interaction strategies. Scientists achieved precise quantitative data regarding the interplay between AI and design thinking.

The study demonstrates that while generative AI doesn’t universally enhance performance, it can demonstrably benefit those with less prior experience. Data shows that effective prompting and iterative refinement are linked to reduced cognitive load, suggesting that strategic engagement with the technology is crucial. These findings indicate a trade-off between openness and controllability in generative models, with Midjourney exhibiting superior image quality, while Stable Diffusion offers greater customizability. Further investigation revealed that generative AI can accelerate visual ideation and broaden formal exploration, expanding a designer’s creative capacity.

The work also highlights that establishing confidence in students has a stronger impact on creativity than simply increasing motivation. Eye-tracking analyses showed shorter visual search paths and more distributed attention when using AI, indicating reduced extraneous cognitive load during design tasks. This research provides valuable insights into how generative AI reshapes the creative process and informs future educational approaches in architectural design.

AI’s Nuance in Early Design Stages

This research investigated the impact of generative AI on conceptual architectural design, examining performance, creative self-efficacy, and cognitive load. Analyses revealed that, overall, utilising generative AI did not significantly alter design performance, cognitive load, or creative self-efficacy when compared to traditional methods in this short-term study. However, the work demonstrates a nuanced effect dependent on user experience, with novice designers exhibiting improved performance during design revision when assisted by generative AI. Notably, the study identified a decline in creative self-efficacy among students employing generative AI, suggesting potential implications for perceived authorship and creative agency.

Furthermore, the research indicates that the manner in which generative AI is used is crucial; iterative refinement and visually-focused prompts correlated with reduced cognitive load during the design process. The authors acknowledge limitations including the absence of post-task qualitative interviews, restricting deeper understanding of participant experiences. Future research should explore the long-term effects of human-AI collaboration, optimal prompt strategies, and the development of design competencies, alongside pedagogically-sound integration of generative AI within architectural education.