Study Reveals Lexical Oscillation in UI/UX Design with Generative AI
A recent study featured on arXiv (2607.23126) investigates how seasoned UI/UX designers establish design intent through natural-language prompts in generative AI tools. The research involved five designers, each with 11-20 years of experience (average 15.4 years), who created landing-page hero sections using a generative AI tool. Data was gathered through think-aloud sessions and retrospective interviews. Employing reflexive thematic analysis with lexical granularity (L1 vibe, L2 design-domain, L3 operational language), it was found that designers do not consistently progress from vague ideas to code; instead, they demonstrate lexical oscillation, often reverting from specific operational language to ambiguity. Discrepancies with AI results led designers to reassess their intent, shifting from instruction to consultation, with one trajectory remaining non-oscillatory.
Key facts
- Study published on arXiv with ID 2607.23126
- Five expert UI/UX designers participated
- Experience range: 11-20 years, mean 15.4 years
- Task: design landing-page hero sections with generative AI
- Data collection: think-aloud and retrospective interviews
- Analysis method: reflexive thematic analysis
- Lexical granularity framework: L1 vibe, L2 design-domain, L3 operational language
- Key finding: lexical oscillation, not unidirectional progression
- Designers returned from operational specificity to ambiguity
- AI output mismatches led to reconsideration of intent
- Engagement shifted from instruction to consultation
- One non-oscillating trajectory observed
Entities
Institutions
- arXiv