Sketching vs. AI Prompt Based Design Intent Evolution in Undergraduate Students: an Exploratory Study

Paper Title

Sketching vs. AI Prompt Based Design Intent Evolution in Undergraduate Students: an Exploratory Study

Publication Info

• Topic area: The impact of AI and traditional sketching on creative processes and outcomes in design education.
• Keywords: AI in design, sketching, creativity, design education, divergent thinking, design intent, cognitive processes, design tools, homogenization, generative AI.

Background and Problem

• Problem / challenge: The introduction of AI tools in early design phases raises concerns about their impact on creative cognitive processes, particularly for novice designers. Existing studies provide mixed evidence on whether AI enhances or limits creativity, and there is insufficient understanding of how AI affects design intent evolution compared to traditional sketching.
• Significance: Understanding how AI influences creativity and cognitive processes is crucial for integrating it effectively into design education and practice, ensuring it enhances rather than detracts from creative exploration.
• Motivation and related work: Prior research has explored AI as a creative support tool, highlighting its potential to enhance divergent thinking but also its risks of homogenization and fixation. Sketching has long been valued for fostering design reasoning and creativity. This study addresses gaps in understanding how AI and sketching compare in supporting creative intent evolution and design space exploration.

Solution

• Proposed approach: An exploratory study comparing sketch-based and AI-based concept design processes among undergraduate industrial design students.
• Novelty:
  1. Empirical comparison of creative outcomes and processes between sketching and AI use in early design phases.
  2. Identification of homogenization effects in AI-generated design concepts.
  3. Analysis of cognitive shifts and design intent evolution in AI-supported versus sketch-based workflows.
  4. Insights into pedagogical implications for integrating AI into design education.
• Procedure and key techniques:
  • Participants: 61 industrial design students divided into sketch (n=31) and AI (n=30) groups.
  • Phases: Diagnostic phase (questionnaire and AUT test), design task phase (water-saving concept ideation), and evaluation phase (creativity assessment using CAT).
  • Data collection: Sketches, AI-generated text/images, and process documentation.
  • Analysis: Creativity ratings, design space exploration patterns, thematic analysis of AI prompting styles, and correlation with AUT scores.

Results

• Concrete findings:
  • Sketch groups generated more image-based ideas (71 ideas, 2.2 per student), while AI groups produced more total ideas (291 ideas, 9.7 per student, including text-based concepts).
  • Creativity ratings were inconclusive: AI groups scored slightly higher on average (2.89 vs. 2.52 on a 5-point scale), but differences were not statistically significant across all comparisons.
  • Homogenization effects were observed in AI groups, with recurring technological solutions lacking contextual diversity.
• Advantage over baselines:
  • Sketching showed clearer design intent evolution and reasoning processes, while AI use expanded the design space but shifted focus toward tool interaction.
• Experiments / evaluation:
  • Creativity was assessed by three design professors using the CAT technique (inter-rater reliability: ICC=0.67).
  • Design space exploration was analyzed through annotated sketches and AI conversation logs.
  • Prompting styles in AI groups were categorized into "jump to the solution," "help me understand the problem," and "cocreate with me."
• Limitations and future work:
  • Limitations include lack of controlled AI model use, no prior training in AI prompting, and reliance on free AI tools with credit restrictions.
  • Future research should explore longitudinal effects, optimal AI integration strategies, and the role of prompting expertise.

Summary

This study compared sketching and AI-based design processes among undergraduate students, focusing on creativity, design intent evolution, and cognitive shifts. While AI expanded the design space and generated more total ideas, sketching supported clearer reasoning and contextual adaptation. Homogenization effects were evident in AI-generated concepts, and creativity ratings were inconclusive. The findings highlight the need for careful integration of AI in design education, emphasizing the importance of traditional skills like sketching and the development of AI tools that foster design reasoning. Future research should address the limitations and explore strategies for optimizing AI use in early design phases.